def __call__(self):
"Return the locations of the ticks"
b = self._transform.base
vmin, vmax = self.axis.get_view_interval()
vmin, vmax = self._transform.transform_point((vmin, vmax))
if vmax < vmin:
vmin, vmax = vmax, vmin
numdec = math.floor(vmax) - math.ceil(vmin)
if self._subs is None:
if numdec > 10:
subs = np.array([1.0])
elif numdec > 6:
subs = np.arange(2.0, b, 2.0)
else:
subs = np.arange(2.0, b)
else:
subs = np.asarray(self._subs)
stride = 1
while numdec / stride + 1 > self.numticks:
stride += 1
decades = np.arange(math.floor(vmin), math.ceil(vmax) + stride, stride)
if len(subs) > 1 or subs[0] != 1.0:
ticklocs = []
for decade in decades:
ticklocs.extend(subs * (np.sign(decade) * b ** np.abs(decade)))
else:
ticklocs = np.sign(decades) * b ** np.abs(decades)
return np.array(ticklocs)
def main():
os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'settings')
from django.core.management import execute_from_command_line
from ot_logbook.models import Activity, Subcategory, Category, Location, \
Equipment, ActivityData
from django.contrib.auth.models import User
argparser = argparse.ArgumentParser(description='fix elevation.')
argparser.add_argument('user', help='user name to bind data')
args = argparser.parse_args()
# Check user
try:
user = User.objects.get(username=args.user)
print 'User : '******'User does not exist.\n')
sys.exit(1)
downloader = SRTMDownloader()
downloader.loadFileList()
activitydata = ActivityData()
elevation_null = ActivityData.objects.filter(elevation__isnull = True)
for row in elevation_null[:2]:
tile = downloader.getTile(int(floor(row.lat)),int(floor(row.lon)))
print "%s Tile lat,lon : %i,%i srtm lat,lon : %f,%f : elevation : %f" % (row.datetime,floor(row.lat),floor(row.lon),row.lat,row.lon,tile.getAltitudeFromLatLon(row.lat,row.lon))
row.elevation = tile.getAltitudeFromLatLon(row.lat,row.lon)
row.save()
print elevation_null.count()
del activitydata
def colorfy(toolname):
print("working")
t = Tool.objects.get(tool_name__iexact=toolname)
if t == None:
return "ERROR - Can't find TOOL!" + toolname
else:
# parse return
numerator = t.quantity_available
denominator = t.quantity
intensity = math.floor(255*1)
try:
n = numerator / denominator
except ZeroDivisionError:
n=0
# now we colorfy it.
r = math.floor(intensity - (intensity*n))
g = math.floor((intensity*n))
b = 0
print(r,g,b)
c = "color: rgb(" + str(r) + "," + str(g) + "," + str(b) +");"
part1 = givemediv(45, c) + str(numerator) + "</div>"
part2 = givemediv(10, "color: black;") + "/" + "</div>"
part3 = givemediv(45, "color: rgb(0, " + str(intensity) + ", 0);" ) + str(denominator) + "</div>"
return part1 + part2 + part3
def downloadChunks(url):
"""Helper to download large files
the only arg is a url
this file will go to a temp directory
the file will also be downloaded
in chunks and print out how much remains
"""
baseFile = path.basename(url)
#move the file to a more uniq path
umask(0002)
try:
temp_path='/tmp'
file = path.join(temp_path,baseFile)
req = urllib2.urlopen(url)
total_size = int(req.info().getheader('Content-Length').strip())
downloaded = 0
CHUNK = 256 * 10240
with open(file, 'wb') as fp:
while True:
chunk = req.read(CHUNK)
downloaded += len(chunk)
print math.floor( (downloaded / total_size) * 100 )
if not chunk: break
fp.write(chunk)
except urllib2.HTTPError, e:
print "HTTP Error:",e.code , url
return False
def calculate_median(self,
minutes=1):
"""
Calculates the median of all values in the collection
after (now - minutes)
"""
now = utc_now()
created_after = now - timedelta(minutes=1)
queryset = self.filter_query(created_after=created_after)
count = queryset.count()
# If there is no value within past 1 minute return None
if count == 0:
return None
# Pick the center element in the collection
# if the collection has a count which is odd
elif (count % 2) != 0:
return queryset.skip(math.floor(count/2))\
.limit(1)\
.all()[0].value
# Pick the center 2 elements in the collection and calculate the avg
# if the collection has a count which is even
elif (count % 2) == 0:
center_elements = queryset.skip(math.floor(count/2) - 1)\
.limit(2)\
.all()
median = (center_elements[0].value + center_elements[1].value)/2.0
return median
def render2(a, b, vol, pos, knum, endamp = .25, sm = 10): b2 = (1. - pause) * b l=waves2(a, b2) ow="" q=int(l[0]*l[1]) lf = log(a) t = (lf-3.) / (8.5-3.) volfac = 1. + .8 * t * cos(pi/5.3*(lf-3.)) snd_len = int((10.-lf)*q) if lf < 4: snd_len *= 2 x = np.arange(snd_len) s = x / float(q) ls = np.log(1. + s) kp_len = int(l[0]) kps1 = np.zeros(snd_len) kps2 = np.zeros(snd_len) kps1[:kp_len] = np.random.normal(size = kp_len) for t in range(kp_len): kps2[t] = kps1[t:t+sm].mean() delt = float(l[0]) li = int(floor(delt)) hi = int(ceil(delt)) ifac = delt % 1 delt2 = delt * (floor(delt) - 1) / floor(delt) ifac2 = delt2 % 1 falloff = (4./lf*endamp)**(1./l[1]) for t in range(hi, snd_len): v1 = ifac * kps2[t-hi] + (1.-ifac) * kps2[t-li] v2 = ifac2 * kps2[t-hi+1] + (1.-ifac2) * kps2[t-li+1] kps2[t] += .5 * (v1 + v2) * falloff data[pos:pos+snd_len] += kps2*vol*volfac
def get_grid_position(x, y, dim_x, dim_y, bounds):
r""" Gives the position of point on a grid
Devides the domain as given by the boundaries into
a grid of dimension :math:`\mathtt{dim_x}\times\mathtt{dim_y}`
and gives the grid-coordinates the point falls into.
Parameters
----------
x, y : float
Coordinates of a point within the domain.
dim_x, dim_y : int
Dimensions of the 2d grid.
bounds : array_like
An array_like object of dimension (2,2)
describing the boundaries of the domain,
e.g. ((min_x, max_x), (min_y, max_x))
Returns
-------
grid_x, grid_y : int
Coordinates of the tile the point falls into.
"""
grid_x = min(math.floor(abs(x - bounds[0][0]) /
abs(bounds[0][0] - bounds[0][1]) * dim_x), dim_x-1)
grid_y = min(math.floor(abs(y - bounds[1][0]) /
abs(bounds[1][0] - bounds[1][1]) * dim_y), dim_y-1)
return (grid_x, grid_y)
def segment_units(self, sfid, verb=False): """ This function takes the list of unit breakpoints, plus the raw metadata, and assembles 'cooked' segments in the corpus segtable. Note: currently ignores the amplitude scalars (aside from generating stats)... """ segmented = self.get_sorted_units_list(sfid) raw_amps, raw_mfccs, raw_chromas = self.get_raw_metadata(sfid) # , raw_chromas amps, reheated = [], [] if verb: print 'raw: ', raw_amps amps_stripped = np.nan_to_num(raw_amps) if verb: print 'amps_stripped: ', amps_stripped mfccs_stripped = np.nan_to_num(raw_mfccs) if verb: print 'mfccs_stripped: ', mfccs_stripped chromas_stripped = np.nan_to_num(raw_chromas) if verb: print 'chromas_stripped: ', chromas_stripped for relid, sfu in enumerate(segmented): offset = int(math.floor(sfu.onset / self.HOP_SECS)) dur = int(math.floor(sfu.dur / self.HOP_SECS)) if verb: print '[[', offset, '|', dur, ']]' self.sftree.nodes[sfid].add_metadata_for_relid(relid, powers=self.feat.powers.proc_funcs[0](amps_stripped, offset, dur)) # WHY ARE THE FUNCTION SIGNATURES DIFFERENT FOR OFFSET AND DUR??? self.sftree.nodes[sfid].add_metadata_for_relid(relid, mfccs=self.feat.proc_funcs[0](mfccs_stripped[offset:(offset+dur)])) if verb: print self.feat.proc_funcs[1] if verb: print mfccs_stripped[offset:(offset+dur)] self.sftree.nodes[sfid].add_metadata_for_relid(relid, mfcc_vars=self.feat.proc_funcs[1](mfccs_stripped[offset:(offset+dur)])) self.sftree.nodes[sfid].add_metadata_for_relid(relid, chromas=self.feat.proc_funcs[0](chromas_stripped[offset:(offset+dur)])) self.sftree.nodes[sfid].add_metadata_for_relid(relid, chroma_vars=self.feat.proc_funcs[1](chromas_stripped[offset:(offset+dur)]))
def bucketSort(array, bucketSize=1):
if len(array) == 0:
return array
# Determine minimum and maximum values
minValue = array[0]
maxValue = array[0]
for i in range(1, len(array)):
if array[i] < minValue:
minValue = array[i]
elif array[i] > maxValue:
maxValue = array[i]
# Initialize buckets
bucketCount = math.floor((maxValue - minValue) / bucketSize) + 1
bucketCount = int(bucketCount)
buckets = []
for i in range(0, bucketCount):
buckets.append([])
# Distribute input array values into buckets
for i in range(0, len(array)):
buckets[int(math.floor((array[i] - minValue) / bucketSize))].append(array[i])
# Sort buckets and place back into input array
array = []
for i in range(0, len(buckets)):
insertSort(buckets[i])
for j in range(0, len(buckets[i])):
array.append(buckets[i][j])
return array
def pos_to_coords(self, pos):
"""Take pixel positions and turn them into coordinates (from 0 to n_rows ** 2 - 1)."""
x, y = pos
coords = (int(math.floor(x/self.tile_line_size)), int(math.floor(y/self.tile_line_size)))
logger.debug('PygameBoard.pos_to_coords')
logger.debug('{}, {}, {}, {}'.format('Coords', coords, type(coords), [type(x) for x in coords]))
return coords
def format_seconds(self, elapsed):
hours = math.floor(elapsed / 3600)
remainder = elapsed % 3600
minutes = math.floor(remainder / 60)
seconds = remainder % 60
return "%.2d:%.2d:%.2d" % (hours, minutes, seconds)
def format(self, record):
'''format a record for display'''
returns = []
line_len = self.width
if type(record.msg) in types.StringTypes:
for line in logging.Formatter.format(self, record).split('\n'):
if len(line) <= line_len:
returns.append(line)
else:
inner_lines = int(math.floor(float(len(line)) / line_len))+1
for i in xrange(inner_lines):
returns.append("%s" % (line[i*line_len:(i+1)*line_len]))
elif type(record.msg) == types.ListType:
if not record.msg:
return ''
# getMessage() must be called so that arguments are substituted; eval() is used to turn the string back into a list
msgdata = eval(record.getMessage())
msgdata.sort()
msgwidth = self.width
columnWidth = max([len(str(item)) for item in msgdata])
columns = int(math.floor(float(msgwidth) / (columnWidth+2)))
lines = int(math.ceil(float(len(msgdata)) / columns))
for lineNumber in xrange(lines):
indices = [idx for idx in [(colNum * lines) + lineNumber
for colNum in range(columns)] if idx < len(msgdata)]
format = (len(indices) * (" %%-%ds " % columnWidth))
returns.append(format % tuple([msgdata[idx] for idx in indices]))
#elif type(record.msg) == lxml.etree._Element:
# returns.append(str(xml_print(record.msg)))
else:
returns.append(str(record.getMessage()))
if record.exc_info:
returns.append(self.formatException(record.exc_info))
return '\n'.join(returns)
def int_to_base_array(x, base): """ Pass the number you wish to format, plus a base number (e.g. for Arabic numerals, 10). You will receive an big-endian array of numbers representing each glyph of your formatted number. Examples: int_to_base_array(255,10) returns [2,5,5] int_to_base_array(255,2) returns [1,1,1,1,1,1,1,1] Don't try base-1! @type x: integer @param x: the integer you wish to convert to a base array @type base: integer @param base: the base-number of the array you wish to convert to @rtype: list @return: A list of integers representing each glyph of your base-X numeral, in big-endian format. """ arr = [] modulo = base while modulo <= x: arr.append(int(math.floor((x % modulo)/(modulo/base)))) modulo = modulo * base arr.append(int(math.floor((x % modulo)/(modulo/base)))) # And then once more arr.reverse() return arr
def findMousePos(self, pos):
# put hit position in window relative coords
x = pos[0] - self.rect[0] - 2
y = pos[1] - self.rect[1] - 5
font_width, font_heigth = getRenderer().getTextSize( ' ', self.font )
col = max( math.floor( (float(x) / float(font_width)) ) + self.scroll[0], 0 )
row = max( math.floor( (float(y) / float(font_heigth)) ) + self.scroll[1], 0 )
r, c, i = 0, 0, 0
while r != row:
try: i = self.text.index( '\n', i ) + 1
except: break
r += 1
while c < col:
if i >= len( self.text) or self.text[i] == '\n':
break
elif self.text[i] == '\t':
if ( c + 4 > col ): break
c += 8
else:
c += 1
i += 1
return min( max( i, 0 ), len( self.text ) )
def divideRect( self, x, y, width, height, c1, c2, c3, c4 ):
new_width = math.floor( width / 2 )
new_height = math.floor( height / 2 )
if ( width > 1 or height > 1 ):
# average of all the points and normalize in case of "out of bounds" during displacement
mid = self.normalize( self.normalize( ( ( c1 + c2 + c3 + c4 ) / 4 ) + self.displace( new_width + new_height ) ) )
# midpoint of the edges is the average of its two end points
edge1 = self.normalize( ( c1 + c2 ) / 2 )
edge2 = self.normalize( ( c2 + c3 ) / 2 )
edge3 = self.normalize( ( c3 + c4 ) / 2 )
edge4 = self.normalize( ( c4 + c1 ) / 2 )
# recursively go down the rabbit hole
self.divideRect( x, y, new_width, new_height, c1, edge1, mid, edge4 )
self.divideRect( x + new_width, y, new_width, new_height, edge1, c2, edge2, mid )
self.divideRect( x + new_width, y + new_height, new_width, new_height, mid, edge2, c3, edge3 )
self.divideRect( x, y + new_height, new_width, new_height, edge4, mid, edge3, c4 )
else:
c = ( c1 + c2 + c3 + c4 ) / 4
self.heightmap[x][y] = c
if ( width == 2 ):
self.heightmap[x + 1][y] = c
if ( height == 2 ):
self.heightmap[x][y + 1] = c
if ( width == 2 and height == 2 ):
self.heightmap[x + 1][y + 1] = c
def SetValue(self, value):
""" Sets the FloatSpin value. """
if not self._textctrl or not self.InRange(value):
return
if self._snapticks and self._increment != 0.0:
finite, snap_value = self.IsFinite(value)
if not finite: # FIXME What To Do About A Failure?
if (snap_value - floor(snap_value) < ceil(snap_value) - snap_value):
value = self._defaultvalue + floor(snap_value)*self._increment
else:
value = self._defaultvalue + ceil(snap_value)*self._increment
strs = ("%100." + str(self._digits) + self._textformat[1])%value
strs = strs.strip()
strs = self.ReplaceDoubleZero(strs)
if value != self._value or strs != self._textctrl.GetValue():
self._textctrl.SetValue(strs)
self._textctrl.DiscardEdits()
self._value = value
def getElementAtPoint(self, point):
'''
Returns a leaf Element which contains the point.
'''
# First, we check that the point does fall inside the mesh.
if not self.boundingBox.containsPoint(point):
return None
# Since the root elements in the mesh have a fixed size and spatial arrangement, it's simple to
# figure out which root element a point is in without having to poll any other elements.
# Start by converting the point in (x,y) in global units into a relative coord (i,j) measured in cell counts.
relativeLocation = (point - self.bottomLeft).scaledBy(1/self.maxCellSize)
i = math.floor(relativeLocation.x)
j = math.floor(relativeLocation.y)
# Figure out which element that is.
e = self.elements[i*self.verticalCellCount+j]
# As a safety net, we check that the element does contain the point. If it doesn't, we risk infinite
# recursion. There's probably something wrong if that happens, so let's raise an exception.
if e.boundingBox.containsPoint(point):
return e.getElementAtPoint(point)
else:
print("Need to query an element",e,"for a point ",point,", but it's the wrong element.")
raise Exception("Fatal Error: Parent mesh attempted to query an element for a point it did not contain.")
def interpolate_cartesian(self, start_pos, start_rot, end_pos, end_rot, pos_spacing, rot_spacing, num_steps = 0):
#normalize quaternion rotations just in case
norm_start_rot = self.normalize_vect(start_rot)
norm_end_rot = self.normalize_vect(end_rot)
#the desired wrist translation
diff = [x-y for (x,y) in zip(end_pos, start_pos)]
if num_steps == 0:
#compute how far the wrist translates
pos_move = self.vect_norm(diff)
#compute how far the wrist rotates
rot_move = self.quat_angle(norm_start_rot, norm_end_rot)
#compute the number of steps to move no more than pos_spacing and rot_spacing in each step
#(min 2, start and end)
num_steps_pos = math.floor(pos_move/pos_spacing)+1
num_steps_rot = math.floor(rot_move/rot_spacing)+1
num_steps = int(max([num_steps_pos, num_steps_rot])+1)
#interpolate
steps = []
for stepind in range(num_steps):
fraction = float(stepind)/(num_steps-1) #add both start (0) and end (1)
rot = list(tf.transformations.quaternion_slerp(norm_start_rot, norm_end_rot, fraction))
pos = list(numpy.array(diff)*fraction + numpy.array(start_pos))
steps.append((pos, rot))
#print "fraction: %5.3f"%fraction, "pos:", pplist(pos), "rot:", pplist(rot)
return steps
def setAppearance(self):
"""
A setter for the appearance of the tower.
"""
self.towerDiv = avg.DivNode(size=util.towerDivSize, pos=(self.pos.x - util.towerDivSize[0]//2, self.pos.y-util.towerDivSize[1]//2))
#sets the explosion radius
self.towerCircle = avg.CircleNode(fillopacity=0.3, strokewidth=0, fillcolor=self.team.color, r=self.towerDiv.size.x//2, pos=(self.towerDiv.size.x//2,self.towerDiv.size.y//2), parent=self.towerDiv)
#sets the fancy snow balls
for i in xrange(5):
radius = self.towerDiv.size[0]//10
xPos = random.randint(0 + math.floor(radius), math.ceil(self.towerDiv.size.x - radius))
yPos = random.randint(0 + math.floor(radius), math.ceil(self.towerDiv.size.y - radius))
snowball = avg.CircleNode(fillopacity=0.5, strokewidth=0, filltexhref=os.path.join(getMediaDir(__file__, "resources"), "snowflakes.png"), r=radius, pos=(xPos,yPos), parent=self.towerDiv)
self.snowballAnim(xPos,yPos,snowball)
self.tower = avg.RectNode(fillopacity=1, strokewidth=0, size=util.towerSize, pos=(self.pos.x - util.towerSize[0] // 2, self.pos.y - util.towerSize[1] // 2))
if self.team.name == "Team2":
self.tower.filltexhref = os.path.join(getMediaDir(__file__, "resources"), "iceball.png")
else:
self.tower.filltexhref = os.path.join(getMediaDir(__file__, "resources"), "iceball.png")
def finish_initializing(self, builder): # pylint: disable=E1002
self.sm = SessionManager('gecos-firstart')
self.sm.start()
iconfile = config.get_data_file('media', '%s' % ('wizard1.png',))
self.set_icon_from_file(iconfile)
screen = Gdk.Screen.get_default()
sw = math.floor(screen.width() - screen.width() / 8)
sh = math.floor(screen.height() - screen.height() / 9)
self.resize(sw, sh)
self.ui.btnTest.set_visible(False)
self.ui.btnTest.set_sensitive(False)
self.show_browser()
self.block()
self.dbusclient = DBusClient()
self.dbusclient.connect('state-changed', self.on_dbusclient_state_changed)
try:
self.dbusclient.start()
self.dbusclient.user_login()
state = self.dbusclient.get_state(reply_handler=self.reply_handler, error_handler=self.error_handler)
except Exception as e:
self.unblock()
def _drawTester(self, dt):
#test box delete
time.sleep(dt)
(r,c,t,s) = self.world.spArray[0]
self.update_idletasks()
self._dbox(r,c)
time.sleep(dt)
(r,c,t,s) = self.world.spArray[4]
self.update_idletasks()
self._dbox(r,c)
time.sleep(dt)
(r,c,t,s) = self.world.spArray[23]
self.update_idletasks()
self._dbox(r,c)
time.sleep(dt)
(r,c,t,s) = self.world.spArray[2]
self.update_idletasks()
self._dbox(r,c)
time.sleep(dt)
(r,c,t,s) = self.world.spArray[12]
self.update_idletasks()
self._dbox(r,c)
time.sleep(dt)
(r,c,t,s) = self.world.spArray[6]
self.update_idletasks()
self._dbox(r,c)
#test cannon ball redraw
for i in range(50):
time.sleep(dt)
(r,c) = (floor(6*i*dt-.5*(i*dt)**2), floor(5*i*dt))
bln = not i%12
self._ushot(r,c,bln)
self._pupdate(15, (float(i)/49)*45)
self.update_idletasks()
print (float(i)/49)*45
def stftFiltering(x, fs, w, N, H, filter): # apply a filter to a sound by using the STFT # x: input sound, w: analysis window, N: FFT size, H: hop size # filter: magnitude response of filter with frequency-magnitude pairs (in dB) # returns y: output sound M = w.size # size of analysis window hM1 = int(math.floor((M+1)/2)) # half analysis window size by rounding hM2 = int(math.floor(M/2)) # half analysis window size by floor x = np.append(np.zeros(hM2),x) # add zeros at beginning to center first window at sample 0 x = np.append(x,np.zeros(hM1)) # add zeros at the end to analyze last sample pin = hM1 # initialize sound pointer in middle of analysis window pend = x.size-hM1 # last sample to start a frame w = w / sum(w) # normalize analysis window y = np.zeros(x.size) # initialize output array while pin<=pend: # while sound pointer is smaller than last sample #-----analysis----- x1 = x[pin-hM1:pin+hM2] # select one frame of input sound mX, pX = DFT.dftAnal(x1, w, N) # compute dft #------transformation----- mY = mX + filter # filter input magnitude spectrum #-----synthesis----- y1 = DFT.dftSynth(mY, pX, M) # compute idft y[pin-hM1:pin+hM2] += H*y1 # overlap-add to generate output sound pin += H # advance sound pointer y = np.delete(y, range(hM2)) # delete half of first window which was added in stftAnal y = np.delete(y, range(y.size-hM1, y.size)) # add zeros at the end to analyze last sample return y
def from_jd(cls, jd):
"""
Convert a Julian day number to a year/month/day tuple
of this calendar (matching jQuery calendars algorithm)
@param jd: the Julian day number
"""
jd = math.floor(jd) + 0.5;
depoch = jd - cls.to_jd(475, 1, 1)
cycle = math.floor(depoch / 1029983)
cyear = math.fmod(depoch, 1029983)
if cyear != 1029982:
aux1 = math.floor(cyear / 366)
aux2 = math.fmod(cyear, 366)
ycycle = math.floor(((2134 * aux1) + (2816 * aux2) + 2815) / 1028522) + aux1 + 1
else:
ycycle = 2820
year = ycycle + (2820 * cycle) + 474
if year <= 0:
year -= 1
yday = jd - cls.to_jd(year, 1, 1) + 1
if yday <= 186:
month = math.ceil(yday / 31)
else:
month = math.ceil((yday - 6) / 30)
day = jd - cls.to_jd(year, month, 1) + 1
return (int(year), int(month), int(day))
def optimize_for_oom(oom):
self.colorbar_step = math.floor(step / 10**oom)*10**oom
self.colorbar_vmin = math.floor(vmin / 10**oom)*10**oom
self.colorbar_vmax = self.colorbar_vmin + \
self.colorbar_step * (number_of_ticks - 1)
self.colorbar_locs = np.arange(0, number_of_ticks
)* self.colorbar_step + self.colorbar_vmin
def plotLimits(self, xy):
ymax = -1.e10
ymin = 1.e10
for x, y in xy:
if y > ymax: ymax = y
if y < ymin: ymin = y
dy = abs(ymax - ymin)
if dy < 0.2*ymin:
ymin = ymin*.9
ymax = ymax*1.1
dy = abs(ymax - ymin)
else:
ymin = ymin - 0.1*dy
ymax = ymax + 0.1*dy
dy = abs(ymax - ymin)
p10 = math.floor(math.log10(0.1*dy))
fctr = math.pow(10.0, p10)
mm = [2.0, 2.5, 2.0]
i = 0
while dy/fctr > 5:
fctr = mm[i % 3]*fctr
i = i + 1
ymin = fctr*math.floor(ymin/fctr)
ymax = fctr*(math.floor(ymax/fctr + 1))
return (ymin, ymax, fctr)
def _write_float(f, x): import math if x < 0: sign = 0x8000 x = x * -1 else: sign = 0 if x == 0: expon = 0 himant = 0 lomant = 0 else: fmant, expon = math.frexp(x) if expon > 16384 or fmant >= 1: # Infinity or NaN expon = sign|0x7FFF himant = 0 lomant = 0 else: # Finite expon = expon + 16382 if expon < 0: # denormalized fmant = math.ldexp(fmant, expon) expon = 0 expon = expon | sign fmant = math.ldexp(fmant, 32) fsmant = math.floor(fmant) himant = long(fsmant) fmant = math.ldexp(fmant - fsmant, 32) fsmant = math.floor(fmant) lomant = long(fsmant) _write_short(f, expon) _write_long(f, himant) _write_long(f, lomant)
def tile(lng, lat, zoom, truncate=False):
"""Get the tile containing a longitude and latitude
Parameters
----------
lng, lat : float
A longitude and latitude pair in decimal degrees.
zoom : int
The web mercator zoom level.
truncate : bool, optional
Whether or not to truncate inputs to limits of web mercator.
Returns
-------
Tile
"""
if truncate:
lng, lat = truncate_lnglat(lng, lat)
lat = math.radians(lat)
n = 2.0 ** zoom
xtile = int(math.floor((lng + 180.0) / 360.0 * n))
try:
ytile = int(math.floor((1.0 - math.log(
math.tan(lat) + (1.0 / math.cos(lat))) / math.pi) / 2.0 * n))
except ValueError:
raise InvalidLatitudeError(
"Y can not be computed for latitude {} radians".format(lat))
else:
return Tile(xtile, ytile, zoom)
def from_context_stroke(self, context): ''' Get the DCS bounds of the path in the graphics context. Stroke, that is, as inked, not as ideal path. ''' # extents of rect in UCS, aligned with axis ulx, uly, lrx, lry = context.stroke_extents() # Two other corners of the rect llx = ulx lly = lry urx = lrx ury = uly # Four points in DCS, corners of rect NOT axis aligned, # and no relationships known between points in DCS p1xd, p1yd = context.user_to_device(ulx, uly) p2xd, p2yd = context.user_to_device(llx, lly) p3xd, p3yd = context.user_to_device(lrx, lry) p4xd, p4yd = context.user_to_device(urx, ury) # DCS bounds are min and max of device coords of all four points. # Snap to outside pixel using floor, ceiling. # Convert to int minxi = int(math.floor(min(p1xd, p3xd, p2xd, p4xd))) minyi = int(math.floor(min(p1yd, p3yd, p2yd, p4yd))) maxxi = int(math.ceil(max(p1xd, p3xd, p2xd, p4xd))) maxyi = int(math.ceil(max(p1yd, p3yd, p2yd, p4yd))) width = maxxi - minxi height = maxyi - minyi self = Bounds(minxi, minyi, width, height) # !!! Cannot assert not is_null: line width tiny or other factors # may yield empty stroke extents. return self
def from_extents(self, ulx, uly, lrx, lry):
'''
Set value from an extent tuple.
For example, cairo extents, floats, inked, converted to DCS.
!!! ulx may be greater than lrx, etc. due to transformations
!!! Extents may be either path (ideal) or stroke (inked).
The ideal extent of a line can have zero width or height.
!!! User may zoom out enough that bounds approach zero,
even equal zero?
!!! Parameters are float i.e. fractional.
Bounds are snapped to the outside pixel boundary.
'''
# Snap to integer boundaries and order on the number line.
# !!! Note int(floor()) not just int()
minxi = int(math.floor(min(ulx, lrx)))
minyi = int(math.floor(min(uly, lry)))
maxxi = int(math.ceil(max(ulx, lrx)))
maxyi = int(math.ceil(max(uly, lry)))
width = maxxi - minxi
height = maxyi - minyi
# width or height or both can be zero, for example setting transform on empty model
# snap float rect to outside integral pixel
## self = gdk.Rectangle(minxi, minyi, width, height)
self = Bounds(minxi, minyi, width, height)
# not assert x,y positive
# assert self.width >= 0 # since abs used
if self.is_null():
print "!!!!!!!!!!!! Null bounds", self
return self
def _hintIter(self, monkey):
angle = 45
dir = 0
dA = 1
pow = self.world.P[1]
x_m = monkey[1]
y_m = monkey[0]
y = x_m*math.tan(angle*(math.pi/180)) - (self.world.g/2) * (x_m/(pow * math.cos(angle*(math.pi/180))))**2
if floor(y) < y_m:
return "Not in range"
while(floor(y) != y_m and angle >= 0 and angle <= 90):
if floor(y) < y_m:
if dir == -1:
dA = dA/2
angle += dA
theta = angle * (math.pi / 180)
y = x_m*math.tan(theta) - self.world.g * 0.5 * (x_m**2) / (pow**2) / ((math.cos(theta))**2)
dir = 1
else:
if dir == 1:
dA = dA/2
angle -= dA
theta = angle * (math.pi / 180)
y = x_m*math.tan(theta) - self.world.g * 0.5 * (x_m**2) / (pow**2) / ((math.cos(theta))**2)
dir = -1
if floor(y) == y_m:
return angle
else:
return "Angle not found"
def calc_change_pct(self, iteracao, max_iter):
ans = math.floor(self.bits * 0.7 * math.floor(
(max_iter - iteracao) / 100.0) / 10.0)
if ans > 1.0:
return float(ans)
return 1.0
def training_worker(graph_manager, task_parameters, user_batch_size,
user_episode_per_rollout):
try:
# initialize graph
graph_manager.create_graph(task_parameters)
# save initial checkpoint
graph_manager.save_checkpoint()
# training loop
steps = 0
graph_manager.setup_memory_backend()
graph_manager.signal_ready()
# To handle SIGTERM
door_man = utils.DoorMan()
# print('---------------- hook.out_dir ----------------')
# print(hook.out_dir)
# print('---------------- hook.dry_run ----------------')
# print(hook.dry_run)
# print('---------------- hook.save_config ----------------')
# print(hook.save_config)
# print('---------------- hook.include_regex ----------------')
# print(hook.include_regex)
# print('---------------- hook.include_collections ----------------')
# print(hook.include_collections)
# print('---------------- hook.save_all ----------------')
# print(hook.save_all)
# print('---------------- hook.include_workers ----------------')
# print(hook.include_workers)
for level in graph_manager.level_managers:
for agent in level.agents.values():
for item in agent.networks.items():
name = item[0]
network = item[1]
print("NETWORK:")
print(name)
print(network)
if network.global_network is not None:
network.global_network.optimizer = graph_manager.smdebug_hook.wrap_optimizer(
network.global_network.optimizer)
if network.online_network is not None:
network.online_network.optimizer = graph_manager.smdebug_hook.wrap_optimizer(
network.online_network.optimizer)
if network.target_network is not None:
network.target_network.optimizer = graph_manager.smdebug_hook.wrap_optimizer(
network.target_network.optimizer)
while steps < graph_manager.improve_steps.num_steps:
# Collect profiler information only IS_PROFILER_ON is true
with utils.Profiler(
s3_bucket=PROFILER_S3_BUCKET,
s3_prefix=PROFILER_S3_PREFIX,
output_local_path=TRAINING_WORKER_PROFILER_PATH,
enable_profiling=IS_PROFILER_ON):
graph_manager.phase = core_types.RunPhase.TRAIN
graph_manager.fetch_from_worker(
graph_manager.agent_params.algorithm.
num_consecutive_playing_steps)
graph_manager.phase = core_types.RunPhase.UNDEFINED
episodes_in_rollout = graph_manager.memory_backend.get_total_episodes_in_rollout(
)
for level in graph_manager.level_managers:
for agent in level.agents.values():
agent.ap.algorithm.num_consecutive_playing_steps.num_steps = episodes_in_rollout
agent.ap.algorithm.num_steps_between_copying_online_weights_to_target.num_steps = episodes_in_rollout
if graph_manager.should_train():
# Make sure we have enough data for the requested batches
rollout_steps = graph_manager.memory_backend.get_rollout_steps(
)
if any(rollout_steps.values()) <= 0:
log_and_exit(
"No rollout data retrieved from the rollout worker",
SIMAPP_TRAINING_WORKER_EXCEPTION,
SIMAPP_EVENT_ERROR_CODE_500)
episode_batch_size = user_batch_size if min(
rollout_steps.values(
)) > user_batch_size else 2**math.floor(
math.log(min(rollout_steps.values()), 2))
# Set the batch size to the closest power of 2 such that we have at least two batches, this prevents coach from crashing
# as batch size less than 2 causes the batch list to become a scalar which causes an exception
for level in graph_manager.level_managers:
for agent in level.agents.values():
agent.ap.network_wrappers[
'main'].batch_size = episode_batch_size
steps += 1
graph_manager.phase = core_types.RunPhase.TRAIN
graph_manager.train()
graph_manager.phase = core_types.RunPhase.UNDEFINED
# Check for Nan's in all agents
rollout_has_nan = False
for level in graph_manager.level_managers:
for agent in level.agents.values():
if np.isnan(agent.loss.get_mean()):
rollout_has_nan = True
if rollout_has_nan:
log_and_exit(
"NaN detected in loss function, aborting training.",
SIMAPP_TRAINING_WORKER_EXCEPTION,
SIMAPP_EVENT_ERROR_CODE_500)
if graph_manager.agent_params.algorithm.distributed_coach_synchronization_type == DistributedCoachSynchronizationType.SYNC:
graph_manager.save_checkpoint()
else:
graph_manager.occasionally_save_checkpoint()
# Clear any data stored in signals that is no longer necessary
graph_manager.reset_internal_state()
for level in graph_manager.level_managers:
for agent in level.agents.values():
agent.ap.algorithm.num_consecutive_playing_steps.num_steps = user_episode_per_rollout
agent.ap.algorithm.num_steps_between_copying_online_weights_to_target.num_steps = user_episode_per_rollout
# for item in agent.networks.items():
# name = item[0]
# network = item[1]
# print("NETWORK:")
# print(name)
# print(network)
# print("--------------------------global_network--------------------------")
# print(network.global_network)
# print("--------------------------online_network--------------------------")
# print(network.online_network)
# print("--------------------------target_network--------------------------")
# print(network.target_network)
# if network.global_network is not None:
# hook.add_to_collection("losses", network.global_network.total_loss)
# smdebug_collection.add(network.global_network.total_loss)
# if network.online_network is not None:
# hook.add_to_collection("losses", network.online_network.total_loss)
# smdebug_collection.add(network.online_network.total_loss)
# if network.target_network is not None:
# hook.add_to_collection("losses", network.target_network.total_loss)
# smdebug_collection.add(network.target_network.total_loss)
if door_man.terminate_now:
log_and_exit(
"Received SIGTERM. Checkpointing before exiting.",
SIMAPP_TRAINING_WORKER_EXCEPTION,
SIMAPP_EVENT_ERROR_CODE_500)
graph_manager.save_checkpoint()
break
except ValueError as err:
if utils.is_error_bad_ckpnt(err):
log_and_exit("User modified model: {}".format(err),
SIMAPP_TRAINING_WORKER_EXCEPTION,
SIMAPP_EVENT_ERROR_CODE_400)
else:
log_and_exit("An error occured while training: {}".format(err),
SIMAPP_TRAINING_WORKER_EXCEPTION,
SIMAPP_EVENT_ERROR_CODE_500)
except Exception as ex:
log_and_exit("An error occured while training: {}".format(ex),
SIMAPP_TRAINING_WORKER_EXCEPTION,
SIMAPP_EVENT_ERROR_CODE_500)
finally:
graph_manager.data_store.upload_finished_file()
def q(n):
p(int(math.floor(n * 100)))
def make_circuit(n:int,f) -> QuantumCircuit:
# circuit begin
input_qubit = QuantumRegister(n,"qc")
classical = ClassicalRegister(n, "qm")
prog = QuantumCircuit(input_qubit, classical)
prog.h(input_qubit[0]) # number=3
prog.h(input_qubit[1]) # number=4
prog.h(input_qubit[2]) # number=5
prog.h(input_qubit[3]) # number=6
prog.h(input_qubit[4]) # number=21
prog.x(input_qubit[2]) # number=26
Zf = build_oracle(n, f)
repeat = floor(sqrt(2 ** n) * pi / 4)
for i in range(1):
prog.append(Zf.to_gate(), [input_qubit[i] for i in range(n)])
prog.h(input_qubit[0]) # number=1
prog.h(input_qubit[1]) # number=2
prog.h(input_qubit[2]) # number=7
prog.h(input_qubit[3]) # number=8
prog.y(input_qubit[3]) # number=25
prog.x(input_qubit[0]) # number=9
prog.h(input_qubit[1]) # number=32
prog.cz(input_qubit[0],input_qubit[1]) # number=33
prog.h(input_qubit[1]) # number=34
prog.cx(input_qubit[0],input_qubit[1]) # number=35
prog.cx(input_qubit[0],input_qubit[1]) # number=38
prog.x(input_qubit[1]) # number=39
prog.cx(input_qubit[0],input_qubit[1]) # number=40
prog.cx(input_qubit[0],input_qubit[1]) # number=37
prog.cx(input_qubit[0],input_qubit[1]) # number=30
prog.cx(input_qubit[0],input_qubit[2]) # number=22
prog.x(input_qubit[2]) # number=23
prog.y(input_qubit[3]) # number=27
prog.cx(input_qubit[0],input_qubit[2]) # number=24
prog.x(input_qubit[3]) # number=12
prog.cx(input_qubit[1],input_qubit[2]) # number=31
if n>=2:
prog.mcu1(pi,input_qubit[1:],input_qubit[0])
prog.x(input_qubit[0]) # number=13
prog.x(input_qubit[1]) # number=14
prog.x(input_qubit[2]) # number=15
prog.x(input_qubit[3]) # number=16
prog.h(input_qubit[0]) # number=17
prog.h(input_qubit[1]) # number=18
prog.h(input_qubit[2]) # number=19
prog.h(input_qubit[3]) # number=20
prog.h(input_qubit[0])
prog.h(input_qubit[1])
prog.h(input_qubit[2])
prog.h(input_qubit[3])
# circuit end
for i in range(n):
prog.measure(input_qubit[i], classical[i])
return prog
def make_circuit(n:int,f) -> QuantumCircuit:
# circuit begin
input_qubit = QuantumRegister(n,"qc")
classical = ClassicalRegister(n, "qm")
prog = QuantumCircuit(input_qubit, classical)
prog.h(input_qubit[0]) # number=3
prog.rx(-1.3603096190043806,input_qubit[2]) # number=28
prog.h(input_qubit[1]) # number=4
prog.h(input_qubit[2]) # number=5
prog.h(input_qubit[3]) # number=6
prog.h(input_qubit[4]) # number=21
Zf = build_oracle(n, f)
repeat = floor(sqrt(2 ** n) * pi / 4)
for i in range(repeat):
prog.append(Zf.to_gate(), [input_qubit[i] for i in range(n)])
prog.h(input_qubit[0]) # number=1
prog.h(input_qubit[1]) # number=2
prog.h(input_qubit[2]) # number=7
prog.h(input_qubit[3]) # number=8
prog.h(input_qubit[3]) # number=34
prog.cz(input_qubit[4],input_qubit[3]) # number=35
prog.h(input_qubit[3]) # number=36
prog.h(input_qubit[0]) # number=38
prog.cz(input_qubit[1],input_qubit[0]) # number=39
prog.h(input_qubit[0]) # number=40
prog.x(input_qubit[0]) # number=32
prog.cx(input_qubit[1],input_qubit[0]) # number=33
prog.cx(input_qubit[0],input_qubit[1]) # number=24
prog.x(input_qubit[1]) # number=25
prog.x(input_qubit[1]) # number=41
prog.cx(input_qubit[0],input_qubit[1]) # number=26
prog.x(input_qubit[2]) # number=11
prog.cx(input_qubit[2],input_qubit[3]) # number=30
prog.x(input_qubit[3]) # number=12
prog.h(input_qubit[2]) # number=42
if n>=2:
prog.mcu1(pi,input_qubit[1:],input_qubit[0])
prog.x(input_qubit[0]) # number=13
prog.x(input_qubit[1]) # number=14
prog.x(input_qubit[2]) # number=15
prog.x(input_qubit[4]) # number=46
prog.x(input_qubit[3]) # number=16
prog.h(input_qubit[0]) # number=17
prog.h(input_qubit[1]) # number=18
prog.cx(input_qubit[0],input_qubit[2]) # number=43
prog.x(input_qubit[2]) # number=44
prog.h(input_qubit[2]) # number=47
prog.cz(input_qubit[0],input_qubit[2]) # number=48
prog.h(input_qubit[2]) # number=49
prog.rx(-1.9697785938008003,input_qubit[1]) # number=37
prog.h(input_qubit[2]) # number=19
prog.h(input_qubit[3]) # number=20
prog.x(input_qubit[1]) # number=22
prog.x(input_qubit[1]) # number=23
# circuit end
for i in range(n):
prog.measure(input_qubit[i], classical[i])
return prog
def weather(arg):
owm = pyowm.OWM(WeatherAPIKey) # You MUST provide a valid API key
fc = owm.three_hours_forecast('Nizhny Novgorod,ru')
f = fc.get_forecast()
get_weather_info = False
day_counter = 0
current_date = ""
cleared = False
temp_Log = ""
Log = ""
for weather in f:
print(weather.get_reference_time('date'))
if ( (arg == "завтра") and (day_counter == 1) and (not cleared)):
Log = ""
cleared = True
if(current_date != str(weather.get_reference_time('date').date()).split('-')[2]):
temp_Log = temp_Log + (str(weather.get_reference_time('date').date()).split('-')[2] + " " + months[str(weather.get_reference_time('date').date()).split('-')[1]] + ":\n")
current_date = str(weather.get_reference_time('date').date()).split('-')[2]
if (str(weather.get_reference_time('date').time()).split(':')[0] == "09"):
temp_Log = temp_Log + ("\t🌝 Утро: ")
get_weather_info = True
if (str(weather.get_reference_time('date').time()).split(':')[0] == "15"):
temp_Log = temp_Log + ("\t🌞 День: ")
get_weather_info = True
if (str(weather.get_reference_time('date').time()).split(':')[0] == "21"):
temp_Log = temp_Log + ("\t🌛 Вечер: ")
get_weather_info = True
day_counter = day_counter + 1
if (str(weather.get_reference_time('date').time()).split(':')[0] == "03"):
temp_Log = temp_Log + ("\t🌚 Ночь: ")
get_weather_info = True
if (get_weather_info):
wind = "{0} {1}".format(weather.get_wind()['speed'], winddirArray[int(math.floor(weather.get_wind()['deg'] + 0.5) % 16)])
humidity = weather.get_humidity()
temp = weather.get_temperature('celsius')['temp']
status = weather.get_status()
try:
weatherResult = "{0}°C {1} {2}, wind {3} m/s, humidity {4}%".format(temp, status,emojisDict[status], wind, humidity)
temp_Log = temp_Log + weatherResult + "\n"
except TypeError:
weatherResult = "{0}°C {1}, wind {2} m/s, humidity {3}%".format(temp, status, wind,humidity)
temp_Log = temp_Log + weatherResult + "\n"
if (str(weather.get_reference_time('date').time()).split(':')[0] == "21"):
Log = Log + temp_Log
temp_Log = ""
if ( (arg == "сегодня" or arg == "") and day_counter == 1):
return Log,""
if (arg == "завтра" and day_counter == 2):
return Log,""
get_weather_info = False
if(arg == "прогноз"):
return Log,""
def check_need_delay():
global times_processed, start_time
up_time = time.time() - start_time
expected_times_processed = math.floor(up_time / PROCESS_INTERVAL)
return times_processed - expected_times_processed > -2
def step_decay(epoch):
lrate = base_lr * math.pow(gamma, math.floor(epoch / 20))
print("Epoch:", epoch, "Learning rate:", lrate)
return lrate
def is_square(num: int):
if int(math.floor(math.sqrt(num))**2) == num:
return True
else:
return False
def add_scatter(
self,
name: str,
data: Union[Dict, DataFrame],
mapping: Dict = {
"x": "x",
"y": "y",
"z": "z",
"c": "c",
"cs": "cs",
"s": "s",
"labels": "labels",
},
colormap: Union[str, Colormap, List[str], List[Colormap]] = "plasma",
shader: str = "sphere",
point_scale: float = 1.0,
max_point_size: float = 100.0,
fog_intensity: float = 0.0,
saturation_limit: Union[float, List[float]] = 0.2,
categorical: Union[bool, List[bool]] = False,
interactive: bool = True,
has_legend: bool = False,
legend_title: Union[str, List[str]] = None,
legend_labels: Union[Dict, List[Dict]] = None,
min_legend_label: Union[str, float, List[str], List[float]] = None,
max_legend_label: Union[str, float, List[str], List[float]] = None,
series_title: Union[str, List[str]] = None,
ondblclick: Union[str, List[str]] = None,
selected_labels: Union[List, List[List]] = None,
label_index: Union[int, List[int]] = 0,
title_index: Union[int, List[int]] = 0,
):
"""Add a scatter layer to the plot.
Arguments:
name (:obj:`str`): The name of the layer
data (:obj:`dict` or :obj:`DataFrame`): A Python dict or Pandas DataFrame containing the data
Keyword Arguments:
mapping (:obj:`dict`, optional): The keys which contain the data in the input dict or the column names in the pandas :obj:`DataFrame`
colormap (:obj:`str`, :obj:`Colormap`, :obj:`List[str]`, or :obj:`List[Colormap]` optional): The name of the colormap (can also be a matplotlib Colormap object). A list when visualizing multiple series
shader (:obj:`str`, optional): The name of the shader to use for the data point visualization
point_scale (:obj:`float`, optional): The relative size of the data points
max_point_size (:obj:`int`, optional): The maximum size of the data points when zooming in
fog_intensity (:obj:`float`, optional): The intensity of the distance fog
saturation_limit (:obj:`float` or :obj:`List[float]`, optional): The minimum saturation to avoid "gray soup". A list when visualizing multiple series
categorical (:obj:`bool` or :obj:`List[bool]`, optional): Whether this scatter layer is categorical. A list when visualizing multiple series
interactive (:obj:`bool`, optional): Whether this scatter layer is interactive
has_legend (:obj:`bool`, optional): Whether or not to draw a legend
legend_title (:obj:`str` or :obj:`List[str]`, optional): The title of the legend. A list when visualizing multiple series
legend_labels (:obj:`Dict` or :obj:`List[Dict]`, optional): A dict mapping values to legend labels. A list when visualizing multiple series
min_legend_label (:obj:`str`, :obj:`float`, :obj:`List[str]` or :obj:`List[float]`, optional): The label used for the miminum value in a ranged (non-categorical) legend. A list when visualizing multiple series
max_legend_label (:obj:`str`, :obj:`float`, :obj:`List[str]` or :obj:`List[float]`, optional): The label used for the maximum value in a ranged (non-categorical) legend. A list when visualizing multiple series
series_title (:obj:`str` or :obj:`List[str]`, optional): The name of the series (used when multiple properites supplied). A list when visualizing multiple series
ondblclick (:obj:`str` or :obj:`List[str]`, optional): A JavaScript snippet that is executed on double-clicking on a data point. A list when visualizing multiple series
selected_labels: (:obj:`Dict` or :obj:`List[Dict]`, optional): A list of label values to show in the selected box. A list when visualizing multiple series
label_index: (:obj:`int` or :obj:`List[int]`, optional): The index of the label value to use as the actual label (when __ is used to specify multiple values). A list when visualizing multiple series
title_index: (:obj:`int` or :obj:`List[int]`, optional): The index of the label value to use as the selected title (when __ is used to specify multiple values). A list when visualizing multiple series
"""
if mapping["z"] not in data:
data[mapping["z"]] = [0] * len(data[mapping["x"]])
if "pandas" in type(data).__module__:
data = data.to_dict("list")
data_c = data[mapping["c"]]
data_cs = data[mapping["c"]] if mapping["cs"] in data else None
# Check whether the color ("c") are strings
if type(data_c[0]) is str:
raise ValueError('Strings are not valid values for "c".')
# In case there are multiple series defined
n_series = 1
if isinstance(data_c[0], Iterable):
n_series = len(data_c)
else:
data_c = [data_c]
if data_cs is not None and not isinstance(data_cs[0], Iterable):
data_cs = [data_cs]
# Make everything a list that isn't one (or a tuple)
colormap = Faerun.make_list(colormap)
saturation_limit = Faerun.make_list(saturation_limit)
categorical = Faerun.make_list(categorical)
legend_title = Faerun.make_list(legend_title)
legend_labels = Faerun.make_list(legend_labels, make_list_list=True)
min_legend_label = Faerun.make_list(min_legend_label)
max_legend_label = Faerun.make_list(max_legend_label)
series_title = Faerun.make_list(series_title)
ondblclick = Faerun.make_list(ondblclick)
selected_labels = Faerun.make_list(selected_labels, make_list_list=True)
label_index = Faerun.make_list(label_index)
title_index = Faerun.make_list(title_index)
# If any argument list is shorter than the number of series,
# repeat the last element
colormap = Faerun.expand_list(colormap, n_series)
saturation_limit = Faerun.expand_list(saturation_limit, n_series)
categorical = Faerun.expand_list(categorical, n_series)
legend_title = Faerun.expand_list(legend_title, n_series, with_none=True)
legend_labels = Faerun.expand_list(legend_labels, n_series, with_none=True)
min_legend_label = Faerun.expand_list(
min_legend_label, n_series, with_none=True
)
max_legend_label = Faerun.expand_list(
max_legend_label, n_series, with_none=True
)
series_title = Faerun.expand_list(series_title, n_series, with_value="Series")
ondblclick = Faerun.expand_list(ondblclick, n_series, with_none=True)
selected_labels = Faerun.expand_list(selected_labels, n_series)
label_index = Faerun.expand_list(label_index, n_series)
title_index = Faerun.expand_list(title_index, n_series)
# # The c and cs values in the data are a special case, as they should
# # never be expanded
# if type(data[mapping["c"]][0]) is not list and prop_len > 1:
# prop_len = 1
# elif:
# prop_len = len(data[mapping["c"]])
legend = [None] * n_series
is_range = [None] * n_series
min_c = [None] * n_series
max_c = [None] * n_series
for s in range(n_series):
min_c[s] = float(min(data_c[s]))
max_c[s] = float(max(data_c[s]))
len_c = len(data_c[s])
if min_legend_label[s] is None:
min_legend_label[s] = min_c[s]
if max_legend_label[s] is None:
max_legend_label[s] = max_c[s]
is_range[s] = False
if legend_title[s] is None:
legend_title[s] = name
# Prepare the legend
legend[s] = []
if has_legend:
legend_values = []
if categorical[s]:
if legend_labels[s]:
legend_values = legend_labels[s]
else:
legend_values = [(i, str(i)) for i in sorted(set(data_c[s]))]
else:
if legend_labels[s]:
legend_labels[s].reverse()
for value, label in legend_labels[s]:
legend_values.append(
[(value - min_c[s]) / (max_c[s] - min_c[s]), label]
)
else:
is_range[s] = True
for i, val in enumerate(np.linspace(1.0, 0.0, 99)):
legend_values.append(
[val, str(data_c[s][int(math.floor(len_c / 100 * i))])]
)
cmap = None
if isinstance(colormap[s], str):
cmap = plt.cm.get_cmap(colormap[s])
else:
cmap = colormap[s]
for value, label in legend_values:
legend[s].append([list(cmap(value)), label])
# Normalize the data to later get the correct colour maps
if not categorical[s]:
data_c[s] = np.array(data_c[s])
data_c[s] = (data_c[s] - min_c[s]) / (max_c[s] - min_c[s])
if mapping["cs"] in data and len(data_cs) > s:
data_cs[s] = np.array(data_cs[s])
min_cs = min(data_cs[s])
max_cs = max(data_cs[s])
# Avoid zero saturation by limiting the lower bound to 0.1
data_cs[s] = 1.0 - np.maximum(
saturation_limit[s],
np.array((data_cs[s] - min_cs) / (max_cs - min_cs)),
)
# Format numbers if parameters are indeed numbers
if isinstance(min_legend_label[s], (int, float)):
min_legend_label[s] = self.legend_number_format.format(
min_legend_label[s]
)
if isinstance(max_legend_label[s], (int, float)):
max_legend_label[s] = self.legend_number_format.format(
max_legend_label[s]
)
data[mapping["c"]] = data_c
if data_cs:
data[mapping["cs"]] = data_cs
self.scatters[name] = {
"name": name,
"shader": shader,
"point_scale": point_scale,
"max_point_size": max_point_size,
"fog_intensity": fog_intensity,
"interactive": interactive,
"categorical": categorical,
"mapping": mapping,
"colormap": colormap,
"has_legend": has_legend,
"legend_title": legend_title,
"legend": legend,
"is_range": is_range,
"min_c": min_c,
"max_c": max_c,
"min_legend_label": min_legend_label,
"max_legend_label": max_legend_label,
"series_title": series_title,
"ondblclick": ondblclick,
"selected_labels": selected_labels,
"label_index": label_index,
"title_index": title_index,
}
self.scatters_data[name] = data
def Q_Learning(Pr_des, eps_unc, N_EPISODES, SHOW_EVERY, LEARN_RATE, DISCOUNT,
EPS_DECAY, epsilon, i_s, pa, energy_pa, pa2ts, pa_s, pa_t,
act_num, possible_acts, possible_next_states, pick_up, delivery,
test_n, m, n):
wb = Workbook()
sheet_name = 'Simulation' + str(test_n + 1)
s1 = wb.add_sheet(sheet_name)
s1.write(1, 0, 'Task-1')
s1.write(6, 0, 'Task-2')
s1.write(11, 0, 'Task-3')
s1.write(16, 0, 'Task-4')
s1.write(0, 1, '# of Hit')
s1.write(0, 2, '# Avg. Reward')
s1.write(0, 6, 'Total Run Time')
s1.write(0, 7, 'Total Avg. Reward')
inx = 0
# Getting the non-accepting TS states
QL_start_time = timeit.default_timer()
EVERY_PATH = []
episode_rewards = []
# Initialize the Q - table (Between -0.01 and 0)
pa_size = []
q_table = []
agent_s = []
hit_count = []
mission_tracker = []
ep_per_task = []
for i in range(len(energy_pa)):
pa_size.append(len(pa[i].g.nodes()))
q_table.append(np.random.rand(m * n, 9) * 0.001 -
0.001) # of states x # of actions
agent_s.append(i_s[i]) # Initialize the agent's location
hit_count.append(0)
mission_tracker.append(0)
ep_per_task.append([])
ep_rewards = []
ep_trajectories_pa = []
agent_upt_i = []
agent_upt = []
for j in range(len(energy_pa)):
for i in range(len(pa[j].g.nodes())):
if pa[j].g.nodes()[i][1] == 0 or str(pa[j].g.nodes(
)[i][0]) == 'r' + str(
pick_up[j]
): #or str(pa[j].g.nodes()[i][0]) == 'r'+str(delivery[j]): # If the mission changes check here
agent_upt_i.append(pa2ts[j][i])
else:
agent_upt_i.append([])
agent_upt.append(agent_upt_i)
for episode in range(N_EPISODES):
# if episode > 8000:
# epsilon = 0
which_pd = np.random.randint(
len(energy_pa)) # randomly chosing the pick_up delivery states
mission_tracker[which_pd] = mission_tracker[which_pd] + 1
hit = []
ep_rew = []
for i in range(len(energy_pa)):
hit.append(0)
ep_traj_pa = [agent_s[which_pd]] # Initialize the episode trajectory
ep_rew = 0 # Initialize the total episode reward
possible_next_states_copy = copy.deepcopy(
possible_next_states[which_pd])
possible_acts_copy = copy.deepcopy(possible_acts[which_pd])
for t_ep in range(ep_len):
k_ep = ep_len - t_ep # Remaning episode time
if hit[which_pd] == 0:
#print(which_pd)
#print(agent_s[which_pd])
if energy_pa[which_pd][agent_s[
which_pd]] == 0: # Raise the 'hit flag' if the mission is achieved
hit[which_pd] = 1 # re-initialize the agent_s to prevent stuck
agent_s[which_pd] = agent_upt[which_pd].index(
pa2ts[which_pd]
[agent_s[which_pd]]) # Reinitiliaze the pa(region, 0)
hit_count[which_pd] = hit_count[which_pd] + 1
#break
en_list = [
energy_pa[which_pd][i]
for i in possible_next_states_copy[agent_s[which_pd]]
] # Energies of the next possible states
not_possible_index = []
ind_minholder = en_list.index(
min(en_list)) #np.argmin(np.array(en_list))
possible_next_states_minholder = possible_next_states_copy[
agent_s[which_pd]][ind_minholder]
possible_acts_minholder = possible_acts_copy[
agent_s[which_pd]][ind_minholder]
for j in range(
len(possible_next_states_copy[agent_s[which_pd]])):
d_max = en_list[j] + 1
i_max = int(math.floor((k_ep - 1 - d_max) / 2))
thr_fun = 0
for i in range(i_max + 1):
thr_fun = thr_fun + np.math.factorial(k_ep) / (
np.math.factorial(k_ep - i) * np.math.factorial(i)
) * eps_unc**i * (1 - eps_unc)**(k_ep - i)
if thr_fun > Pr_des or i_max < 0: #energy_pa[possible_next_states_copy[agent_s][j]] > k_ep-2: #
not_possible_index.append(j)
#print("r1pos = " + str(len(possible_next_states_copy[agent_s[which_pd]])))
for ind in sorted(not_possible_index, reverse=True):
del possible_next_states_copy[agent_s[which_pd]][ind]
del possible_acts_copy[agent_s[which_pd]][ind]
#print("r2pos = " + str(len(possible_next_states_copy[agent_s[which_pd]])))
if len(possible_next_states_copy[agent_s[which_pd]]
) == 0: # not possible_next_states_copy[agent_s]: #
possible_next_states_copy[agent_s[which_pd]].append(
possible_next_states_minholder)
possible_acts_copy[agent_s[which_pd]].append(
possible_acts_minholder)
if np.random.uniform() > epsilon: # Exploit
possible_qs = q_table[which_pd][pa2ts[which_pd][
agent_s[which_pd]], possible_acts_copy[agent_s[
which_pd]]] # Possible Q values for each action
next_ind = np.argmax(
possible_qs) # Pick the action with max Q value
else: # Explore
next_ind = np.random.randint(
len(possible_acts_copy[
agent_s[which_pd]])) # Picking a random action
# Taking the action
real_agent_s = agent_s[which_pd]
real_phy_agent = pa2ts[which_pd][agent_s[which_pd]]
real_action = possible_acts_copy[real_agent_s][next_ind]
if np.random.uniform() < eps_unc:
[chosen_act, next_state] = action_uncertainity(
possible_acts_copy[agent_s[which_pd]][next_ind],
pa_s[which_pd], pa_t[which_pd], act_num[which_pd],
agent_s[which_pd])
action = chosen_act
s_a = (agent_s[which_pd], action) # State & Action pair
#current_q = q_table[which_pd][agent_s[which_pd], action] # (save the current q for the q_table update later on)
agent_s[
which_pd] = next_state # possible_next_states[agent_upt.index(pa2ts[agent_s])][next_ind] # moving to next state (s,a)
else:
action = possible_acts_copy[agent_s[which_pd]][next_ind]
s_a = (agent_s[which_pd], action) # State & Action pair
#current_q = q_table[which_pd][agent_s[which_pd], action] # (save the current q for the q_table update later on)
agent_s[which_pd] = possible_next_states_copy[
agent_s[which_pd]][next_ind] # moving to next state (s,a)
ep_traj_pa.append(agent_s[which_pd])
current_q = q_table[which_pd][real_phy_agent, real_action]
max_future_q = np.amax(
q_table[which_pd][pa2ts[which_pd][agent_s[which_pd]], :]
) # Find the max future q
rew_obs = rewards_pa[which_pd][agent_s[
which_pd]] #np.random.binomial(1, 1-rew_uncertainity) * # Observe the rewards of the next state
new_q = (1 - LEARN_RATE) * current_q + LEARN_RATE * (
rew_obs + DISCOUNT * max_future_q)
q_table[which_pd][real_phy_agent, real_action] = new_q
# print("episode = " + str(episode))
for i in range(len(energy_pa)):
if i != which_pd:
hypo_agent_s = agent_s[i]
hypo_phy_agent = pa2ts[i][agent_s[i]]
agent_s[i] = agent_upt[i].index(
pa2ts[which_pd][agent_s[which_pd]])
current_q = q_table[i][hypo_phy_agent, real_action]
max_future_q = np.amax(q_table[i][pa2ts[i][agent_s[i]], :])
new_q = (1 - LEARN_RATE) * current_q + LEARN_RATE * (
rew_obs + DISCOUNT * max_future_q
) # Calculate the new q value
q_table[i][hypo_phy_agent,
real_action] = new_q # Update the table
ep_rew += rew_obs
agent_s[which_pd] = agent_upt[which_pd].index(
pa2ts[which_pd][agent_s[which_pd]])
# for i in range(len(energy_pa)):
# agent_s[i] = agent_upt[i].index(pa2ts[i][agent_s[i]])# Reinitiliaze the pa(region, 0)
ep_rewards.append(ep_rew)
ep_trajectories_pa.append(ep_traj_pa)
epsilon = epsilon * EPS_DECAY
ep_per_task[which_pd].append(ep_rew)
if (episode + 1) % 10000 == 0:
inx = inx + 1
for ind in range(len(energy_pa)):
avg_per_task = np.mean(ep_per_task[ind])
print('Episode # ' + str(episode + 1) + ' : Task-' + str(ind) +
' # of Hit=' + str(len(ep_per_task[ind])) +
' Avg.=' + str(avg_per_task))
s1.write(ind * N_EPISODES / 10000 + inx, 1,
len(ep_per_task[ind]))
s1.write(ind * N_EPISODES / 10000 + inx, 2, avg_per_task)
if (episode + 1) % SHOW_EVERY == 0:
avg_rewards = np.mean(ep_rewards[episode - SHOW_EVERY + 1:episode])
print('Episode # ' + str(episode + 1) + ' : Epsilon=' +
str(round(epsilon, 4)) + ' Avg. reward in the last ' +
str(SHOW_EVERY) + ' episodes=' + str(round(avg_rewards, 2)))
best_episode_index = ep_rewards.index(max(ep_rewards))
optimal_policy_pa = ep_trajectories_pa[
N_EPISODES -
1] #ep_trajectories_pa[best_episode_index] # Optimal policy in pa ep_trajectories_pa[N_EPISODES-1]#
optimal_policy_ts = [] # optimal policy in ts
opt_pol = [] # optimal policy in (m, n, h) format for visualization
for ind, val in enumerate(optimal_policy_pa):
optimal_policy_ts.append(pa2ts[which_pd][val])
opt_pol.append((math.floor(optimal_policy_ts[ind] / n),
optimal_policy_ts[ind] % n, 0))
print('\n Tajectory at the last episode : ' + str(optimal_policy_ts))
QL_timecost = timeit.default_timer() - QL_start_time
success_ratio = []
for i in range(len(energy_pa)):
success_ratio.append(100 * hit_count[i] / mission_tracker[i])
print("Successful Mission Ratio[%] = " + str(success_ratio[i]))
print("Successful Missions = " + str(hit_count[i]) + " out of " +
str(mission_tracker[i]))
d_maxs = []
for i in range(len(energy_pa)):
d_maxs.append(max(energy_pa[i]))
max_energy = max(d_maxs)
print('\n Total time for Q-Learning : ' + str(QL_timecost) + ' seconds' +
"\n")
print('Action uncertainity[%] = ' + str(eps_unc * 100))
print("Desired Minimum Success Ratio[%] = " + str(100 * Pr_des))
print("Episode Length = " + str(ep_len) +
" and Max. Energy of the System = " + str(max_energy) + "\n")
s1.write(1, 6, QL_timecost)
s1.write(1, 7, np.mean(ep_rewards))
#filename = 'testing'+str(test_n+1) +'.xls'
filename = 'testing_million_stable_one_mission.xls'
wb.save(filename)
return opt_pol
def get_msdnet_cifar10(blocks,
model_name=None,
pretrained=False,
root=os.path.join('~', '.torch', 'models'),
**kwargs):
"""
Create MSDNet model for CIFAR-10 with specific parameters.
Parameters:
----------
blocks : int
Number of blocks.
model_name : str or None, default None
Model name for loading pretrained model.
pretrained : bool, default False
Whether to load the pretrained weights for model.
root : str, default '~/.torch/models'
Location for keeping the model parameters.
"""
assert (blocks == 22)
num_scales = 3
num_feature_blocks = 10
base = 4
step = 2
reduction_rate = 0.5
growth = 6
growth_factor = [1, 2, 4, 4]
use_bottleneck = True
bottleneck_factor_per_scales = [1, 2, 4, 4]
assert (reduction_rate > 0.0)
init_layer_channels = [16 * c for c in growth_factor[:num_scales]]
step_mode = "even"
layers_per_subnets = [base]
for i in range(num_feature_blocks - 1):
layers_per_subnets.append(step if step_mode == 'even' else step * i +
1)
total_layers = sum(layers_per_subnets)
interval = math.ceil(total_layers / num_scales)
global_layer_ind = 0
channels = []
bottleneck_factors = []
in_channels_tmp = init_layer_channels
in_scales = num_scales
for i in range(num_feature_blocks):
layers_per_subnet = layers_per_subnets[i]
scales_i = []
channels_i = []
bottleneck_factors_i = []
for j in range(layers_per_subnet):
out_scales = int(num_scales -
math.floor(global_layer_ind / interval))
global_layer_ind += 1
scales_i += [out_scales]
scale_offset = num_scales - out_scales
in_dec_scales = num_scales - len(in_channels_tmp)
out_channels = [
in_channels_tmp[scale_offset - in_dec_scales + k] +
growth * growth_factor[scale_offset + k]
for k in range(out_scales)
]
in_dec_scales = num_scales - len(in_channels_tmp)
bottleneck_factors_ij = bottleneck_factor_per_scales[
in_dec_scales:][:len(in_channels_tmp)]
in_channels_tmp = out_channels
channels_i += [out_channels]
bottleneck_factors_i += [bottleneck_factors_ij]
if in_scales > out_scales:
assert (in_channels_tmp[0] % growth_factor[scale_offset] == 0)
out_channels1 = int(
math.floor(in_channels_tmp[0] /
growth_factor[scale_offset] * reduction_rate))
out_channels = [
out_channels1 * growth_factor[scale_offset + k]
for k in range(out_scales)
]
in_channels_tmp = out_channels
channels_i += [out_channels]
bottleneck_factors_i += [[]]
in_scales = out_scales
in_scales = scales_i[-1]
channels += [channels_i]
bottleneck_factors += [bottleneck_factors_i]
net = CIFAR10MSDNet(channels=channels,
init_layer_channels=init_layer_channels,
num_feature_blocks=num_feature_blocks,
use_bottleneck=use_bottleneck,
bottleneck_factors=bottleneck_factors,
**kwargs)
if pretrained:
if (model_name is None) or (not model_name):
raise ValueError(
"Parameter `model_name` should be properly initialized for loading pretrained model."
)
from .model_store import download_model
download_model(net=net,
model_name=model_name,
local_model_store_dir_path=root)
return net
def update_flags(flags):
"""Update flags with new parameters.
Args:
flags: All model and data parameters
Returns:
Updated flags
Raises:
ValueError: If the preprocessing mode isn't recognized.
"""
label_count = len(
input_data.prepare_words_list(flags.wanted_words.split(',')))
desired_samples = int(flags.sample_rate * flags.clip_duration_ms /
MS_PER_SECOND)
window_size_samples = int(flags.sample_rate * flags.window_size_ms /
MS_PER_SECOND)
window_stride_samples = int(flags.sample_rate * flags.window_stride_ms /
MS_PER_SECOND)
length_minus_window = (desired_samples - window_size_samples)
if length_minus_window < 0:
spectrogram_length = 0
else:
spectrogram_length = 1 + int(length_minus_window / window_stride_samples)
if flags.preprocess == 'raw':
average_window_width = -1
fingerprint_width = desired_samples
spectrogram_length = 1
elif flags.preprocess == 'average':
fft_bin_count = 1 + (utils.next_power_of_two(window_size_samples) / 2)
average_window_width = int(
math.floor(fft_bin_count / flags.feature_bin_count))
fingerprint_width = int(
math.ceil(float(fft_bin_count) / average_window_width))
elif flags.preprocess == 'mfcc':
average_window_width = -1
fingerprint_width = flags.feature_bin_count
elif flags.preprocess == 'micro':
average_window_width = -1
fingerprint_width = flags.feature_bin_count
else:
raise ValueError('Unknown preprocess mode "%s" (should be "mfcc",'
' "average", or "micro")' % (flags.preprocess))
fingerprint_size = fingerprint_width * spectrogram_length
upd_flags = flags
upd_flags.mode = Modes.TRAINING
upd_flags.label_count = label_count
upd_flags.desired_samples = desired_samples
upd_flags.window_size_samples = window_size_samples
upd_flags.window_stride_samples = window_stride_samples
upd_flags.spectrogram_length = spectrogram_length
upd_flags.fingerprint_width = fingerprint_width
upd_flags.fingerprint_size = fingerprint_size
upd_flags.average_window_width = average_window_width
# summary logs for TensorBoard
upd_flags.summaries_dir = os.path.join(flags.train_dir, 'logs/')
return upd_flags
def page(self, number, *args, **kwargs):
"""Return a standard ``Page`` instance with custom, digg-specific
page ranges attached.
"""
page = super(DiggPaginator, self).page(number, *args, **kwargs)
number = int(number) # we know this will work
# easier access
num_pages, body, tail, padding, margin = \
self.num_pages, self.body, self.tail, self.padding, self.margin
# put active page in middle of main range
main_range = map(
int,
[
math.floor(number - body / 2.0) +
1, # +1 = shift odd body to right
math.floor(number + body / 2.0)
])
# adjust bounds
if main_range[0] < 1:
main_range = map(abs(main_range[0] - 1).__add__, main_range)
if main_range[1] > num_pages:
main_range = map((num_pages - main_range[1]).__add__, main_range)
# Determine leading and trailing ranges; if possible and appropriate,
# combine them with the main range, in which case the resulting main
# block might end up considerable larger than requested. While we
# can't guarantee the exact size in those cases, we can at least try
# to come as close as possible: we can reduce the other boundary to
# max padding, instead of using half the body size, which would
# otherwise be the case. If the padding is large enough, this will
# of course have no effect.
# Example:
# total pages=100, page=4, body=5, (default padding=2)
# 1 2 3 [4] 5 6 ... 99 100
# total pages=100, page=4, body=5, padding=1
# 1 2 3 [4] 5 ... 99 100
# If it were not for this adjustment, both cases would result in the
# first output, regardless of the padding value.
if main_range[0] <= tail + margin:
leading = []
main_range = [1, max(body, min(number + padding, main_range[1]))]
main_range[0] = 1
else:
leading = range(1, tail + 1)
# basically same for trailing range, but not in ``left_align`` mode
if self.align_left:
trailing = []
else:
if main_range[1] >= num_pages - (tail + margin) + 1:
trailing = []
if not leading:
# ... but handle the special case of neither leading nor
# trailing ranges; otherwise, we would now modify the
# main range low bound, which we just set in the previous
# section, again.
main_range = [1, num_pages]
else:
main_range = [
min(num_pages - body + 1,
max(number - padding, main_range[0])), num_pages
]
else:
trailing = range(num_pages - tail + 1, num_pages + 1)
# finally, normalize values that are out of bound; this basically
# fixes all the things the above code screwed up in the simple case
# of few enough pages where one range would suffice.
main_range = [max(main_range[0], 1), min(main_range[1], num_pages)]
# make the result of our calculations available as custom ranges
# on the ``Page`` instance.
page.main_range = range(main_range[0], main_range[1] + 1)
page.leading_range = leading
page.trailing_range = trailing
page.page_range = reduce(
lambda x, y: x + ((x and y) and [False]) + y,
[page.leading_range, page.main_range, page.trailing_range])
page.__class__ = DiggPage
return page
async def create_filter(self, flags, ctx, order_by=None):
aggregations = []
if "mine" in flags and flags["mine"]:
aggregations.append({"$match": {"user_id": ctx.author.id}})
if "bids" in flags and flags["bids"]:
aggregations.append({"$match": {"bidder_id": ctx.author.id}})
rarity = []
for x in ("mythical", "legendary", "ub"):
if x in flags and flags[x]:
rarity += getattr(self.bot.data, f"list_{x}")
if rarity:
aggregations.append({"$match": {"pokemon.species_id": {"$in": rarity}}})
for x in ("alolan", "mega", "event"):
if x in flags and flags[x]:
aggregations.append(
{"$match": {"pokemon.species_id": {"$in": getattr(self.bot.data, f"list_{x}")}}}
)
if "type" in flags and flags["type"]:
all_species = [i for x in flags["type"] for i in self.bot.data.list_type(x)]
aggregations.append({"$match": {"pokemon.species_id": {"$in": all_species}}})
if "region" in flags and flags["region"]:
all_species = [i for x in flags["region"] for i in self.bot.data.list_region(x)]
aggregations.append({"$match": {"pokemon.species_id": {"$in": all_species}}})
if "favorite" in flags and flags["favorite"]:
aggregations.append({"$match": {"pokemon.favorite": True}})
if "shiny" in flags and flags["shiny"]:
aggregations.append({"$match": {"pokemon.shiny": True}})
if "name" in flags and flags["name"] is not None:
all_species = [
i for x in flags["name"] for i in self.bot.data.find_all_matches(" ".join(x))
]
aggregations.append({"$match": {"pokemon.species_id": {"$in": all_species}}})
if "nickname" in flags and flags["nickname"] is not None:
aggregations.append(
{
"$match": {
"pokemon.nickname": {
"$regex": "("
+ ")|(".join(" ".join(x) for x in flags["nickname"])
+ ")",
"$options": "i",
}
}
}
)
if "embedcolor" in flags and flags["embedcolor"]:
aggregations.append({"$match": {"pokemon.has_color": True}})
if "ends" in flags and flags["ends"] is not None:
aggregations.append({"$match": {"ends": {"$lt": datetime.utcnow() + flags["ends"]}}})
# Numerical flags
for flag, expr in constants.FILTER_BY_NUMERICAL.items():
for text in flags[flag] or []:
ops = self.parse_numerical_flag(text)
if ops is None:
raise commands.BadArgument(f"Couldn't parse `--{flag} {' '.join(text)}`")
ops[1] = float(ops[1])
if flag == "iv":
ops[1] = float(ops[1]) * 186 / 100
if ops[0] == "<":
aggregations.append(
{"$match": {expr: {"$lt": math.ceil(ops[1])}}},
)
elif ops[0] == "=":
aggregations.append(
{"$match": {expr: {"$eq": round(ops[1])}}},
)
elif ops[0] == ">":
aggregations.append(
{"$match": {expr: {"$gt": math.floor(ops[1])}}},
)
for flag, amt in constants.FILTER_BY_DUPLICATES.items():
if flag in flags and flags[flag] is not None:
iv = int(flags[flag])
# Processing combinations
combinations = [
{field: iv for field in combo}
for combo in itertools.combinations(constants.IV_FIELDS, amt)
]
aggregations.append({"$match": {"$or": combinations}})
if order_by is not None:
s = order_by[-1]
if order_by[-1] in "+-":
order_by, asc = order_by[:-1], 1 if s == "+" else -1
else:
asc = -1 if order_by in constants.DEFAULT_DESCENDING else 1
aggregations.append({"$sort": {constants.SORTING_FUNCTIONS[order_by]: asc}})
if "skip" in flags and flags["skip"] is not None:
aggregations.append({"$skip": flags["skip"]})
if "limit" in flags and flags["limit"] is not None:
aggregations.append({"$limit": flags["limit"]})
return aggregations
##### System Inputs for Q-Learning #####
N_EPISODES = 100000 # of episodes
SHOW_EVERY = 10000 # Print out the info at every ... episode
LEARN_RATE = 0.1
DISCOUNT = 0.95
EPS_DECAY = 1 #0.99989
epsilon = 0.1 # 0.95
eps_unc = 0.10 # Uncertainity in actions
Pr_des = 0.95
# Check possible minimum threshold
d_maxs = []
for i in range(len(energy_pa)):
d_maxs.append(max(energy_pa[i]))
d_max = max(d_maxs)
i_max = int(math.floor((ep_len - 1 - d_max) / 2))
thr_fun = 0
for i in range(i_max + 1):
thr_fun = thr_fun + np.math.factorial(ep_len) / (
np.math.factorial(ep_len - i) *
np.math.factorial(i)) * eps_unc**i * (1 - eps_unc)**(ep_len - i)
print('Maximum threshold that can be put ' + str(thr_fun))
if thr_fun < Pr_des:
print('Please set a less desired probability threshold than ' +
str(thr_fun))
else:
# Call the Q_Learning Function
for test_n in range(1):
opt_pol = Q_Learning(Pr_des, eps_unc, N_EPISODES, SHOW_EVERY,
LEARN_RATE, DISCOUNT, EPS_DECAY, epsilon, i_s,
pa, energy_pa, pa2ts, pa_s, pa_t, act_num,
a=1.2 b=2.0 print a.is_integer(),b.is_integer() import math print round(a), math.ceil(a), math.floor(a)
curs.init_pair(1, 7, -1)
curs.init_pair(2, 0, 3)
curs.init_pair(3, 3, -1)
select_idx = 0
key = 0
key_moveup = 106
key_movedown = 107
key_enter = 10
key_q = 113
key_v = 118
key_d = 100
key_u = 117
key_e = 101
lines_mid = math.floor(LINES/2)
title = "P-R-O-J-E-C-T-S"
def move(dire, num, wrap):
wrapLen = len(wrap)
if dire == 'up':
num = (num + 1)%wrapLen
elif dire == 'down':
num = (num - 1)%wrapLen
return num
def select(selected, cords, string):
if selected:
win.scr.addstr(cords[0], cords[1], string, curs.color_pair(2))
else:
def forward(self, # type: ignore
text: Dict[str, torch.LongTensor],
spans: torch.IntTensor,
span_labels: torch.IntTensor = None,
metadata: List[Dict[str, Any]] = None) -> Dict[str, torch.Tensor]:
# pylint: disable=arguments-differ
"""
Parameters
----------
text : ``Dict[str, torch.LongTensor]``, required.
The output of a ``TextField`` representing the text of
the document.
spans : ``torch.IntTensor``, required.
A tensor of shape (batch_size, num_spans, 2), representing the inclusive start and end
indices of candidate spans for mentions. Comes from a ``ListField[SpanField]`` of
indices into the text of the document.
span_labels : ``torch.IntTensor``, optional (default = None).
A tensor of shape (batch_size, num_spans), representing the cluster ids
of each span, or -1 for those which do not appear in any clusters.
metadata : ``List[Dict[str, Any]]``, optional (default = None).
A metadata dictionary for each instance in the batch. We use the "original_text" and "clusters" keys
from this dictionary, which respectively have the original text and the annotated gold coreference
clusters for that instance.
Returns
-------
An output dictionary consisting of:
top_spans : ``torch.IntTensor``
A tensor of shape ``(batch_size, num_spans_to_keep, 2)`` representing
the start and end word indices of the top spans that survived the pruning stage.
antecedent_indices : ``torch.IntTensor``
A tensor of shape ``(num_spans_to_keep, max_antecedents)`` representing for each top span
the index (with respect to top_spans) of the possible antecedents the model considered.
predicted_antecedents : ``torch.IntTensor``
A tensor of shape ``(batch_size, num_spans_to_keep)`` representing, for each top span, the
index (with respect to antecedent_indices) of the most likely antecedent. -1 means there
was no predicted link.
loss : ``torch.FloatTensor``, optional
A scalar loss to be optimised.
"""
import pdb; pdb.set_trace()
# Shape: (batch_size, document_length, embedding_size)
if len(text) <= 512:
text_embeddings = self._lexical_dropout(self._text_field_embedder(text))
else:
# sliding window approach
final_representations = []
current_text = self._text_field_embedder(text[0])
final_representations.append(current_text[:412])
for text_fragment in text[1:]:
next_text = self._text_field_embedder(text_fragment)
# average the 100.
averaged = torch.mean(torch.stack([current_text[412:], next_text[:100]]), dim=0)
final_representations = torch.cat([final_representations, averaged])
current_text = next_text
final_representations = torch.cat([final_representations, current_text[-100:]])
text_embeddings = final_representations # should be [num_sentence, dim_size]
document_length = text_embeddings.size(1)
num_spans = spans.size(1)
# Shape: (batch_size, document_length)
text_mask = util.get_text_field_mask(text).float()
# Shape: (batch_size, num_spans)
span_mask = (spans[:, :, 0] >= 0).squeeze(-1).float()
# SpanFields return -1 when they are used as padding. As we do
# some comparisons based on span widths when we attend over the
# span representations that we generate from these indices, we
# need them to be <= 0. This is only relevant in edge cases where
# the number of spans we consider after the pruning stage is >= the
# total number of spans, because in this case, it is possible we might
# consider a masked span.
# Shape: (batch_size, num_spans, 2)
spans = F.relu(spans.float()).long()
# Shape: (batch_size, document_length, encoding_dim)
contextualized_embeddings = self._context_layer(text_embeddings, text_mask)
# Shape: (batch_size, num_spans, 2 * encoding_dim + feature_size)
endpoint_span_embeddings = self._endpoint_span_extractor(contextualized_embeddings, spans)
# Shape: (batch_size, num_spans, emebedding_size)
attended_span_embeddings = self._attentive_span_extractor(text_embeddings, spans)
# Shape: (batch_size, num_spans, emebedding_size + 2 * encoding_dim + feature_size)
span_embeddings = torch.cat([endpoint_span_embeddings, attended_span_embeddings], -1)
# Prune based on mention scores.
num_spans_to_keep = int(math.floor(self._spans_per_word * document_length))
(top_span_embeddings, top_span_mask,
top_span_indices, top_span_mention_scores) = self._mention_pruner(span_embeddings,
span_mask,
num_spans_to_keep)
top_span_mask = top_span_mask.unsqueeze(-1)
# Shape: (batch_size * num_spans_to_keep)
# torch.index_select only accepts 1D indices, but here
# we need to select spans for each element in the batch.
# This reformats the indices to take into account their
# index into the batch. We precompute this here to make
# the multiple calls to util.batched_index_select below more efficient.
flat_top_span_indices = util.flatten_and_batch_shift_indices(top_span_indices, num_spans)
# Compute final predictions for which spans to consider as mentions.
# Shape: (batch_size, num_spans_to_keep, 2)
top_spans = util.batched_index_select(spans,
top_span_indices,
flat_top_span_indices)
# Compute indices for antecedent spans to consider.
max_antecedents = min(self._max_antecedents, num_spans_to_keep)
# Now that we have our variables in terms of num_spans_to_keep, we need to
# compare span pairs to decide each span's antecedent. Each span can only
# have prior spans as antecedents, and we only consider up to max_antecedents
# prior spans. So the first thing we do is construct a matrix mapping a span's
# index to the indices of its allowed antecedents. Note that this is independent
# of the batch dimension - it's just a function of the span's position in
# top_spans. The spans are in document order, so we can just use the relative
# index of the spans to know which other spans are allowed antecedents.
# Once we have this matrix, we reformat our variables again to get embeddings
# for all valid antecedents for each span. This gives us variables with shapes
# like (batch_size, num_spans_to_keep, max_antecedents, embedding_size), which
# we can use to make coreference decisions between valid span pairs.
# Shapes:
# (num_spans_to_keep, max_antecedents),
# (1, max_antecedents),d
# (1, num_spans_to_keep, max_antecedents)
valid_antecedent_indices, valid_antecedent_offsets, valid_antecedent_log_mask = \
self._generate_valid_antecedents(num_spans_to_keep, max_antecedents, util.get_device_of(text_mask))
# Select tensors relating to the antecedent spans.
# Shape: (batch_size, num_spans_to_keep, max_antecedents, embedding_size)
candidate_antecedent_embeddings = util.flattened_index_select(top_span_embeddings,
valid_antecedent_indices)
# Shape: (batch_size, num_spans_to_keep, max_antecedents)
candidate_antecedent_mention_scores = util.flattened_index_select(top_span_mention_scores,
valid_antecedent_indices).squeeze(-1)
# Compute antecedent scores.
# Shape: (batch_size, num_spans_to_keep, max_antecedents, embedding_size)
span_pair_embeddings = self._compute_span_pair_embeddings(top_span_embeddings,
candidate_antecedent_embeddings,
valid_antecedent_offsets)
# Shape: (batch_size, num_spans_to_keep, 1 + max_antecedents)
coreference_scores = self._compute_coreference_scores(span_pair_embeddings,
top_span_mention_scores,
candidate_antecedent_mention_scores,
valid_antecedent_log_mask)
# We now have, for each span which survived the pruning stage,
# a predicted antecedent. This implies a clustering if we group
# mentions which refer to each other in a chain.
# Shape: (batch_size, num_spans_to_keep)
_, predicted_antecedents = coreference_scores.max(2)
# Subtract one here because index 0 is the "no antecedent" class,
# so this makes the indices line up with actual spans if the prediction
# is greater than -1.
predicted_antecedents -= 1
output_dict = {"top_spans": top_spans,
"antecedent_indices": valid_antecedent_indices,
"predicted_antecedents": predicted_antecedents}
if span_labels is not None:
# Find the gold labels for the spans which we kept.
pruned_gold_labels = util.batched_index_select(span_labels.unsqueeze(-1),
top_span_indices,
flat_top_span_indices)
antecedent_labels = util.flattened_index_select(pruned_gold_labels,
valid_antecedent_indices).squeeze(-1)
antecedent_labels += valid_antecedent_log_mask.long()
# Compute labels.
# Shape: (batch_size, num_spans_to_keep, max_antecedents + 1)
gold_antecedent_labels = self._compute_antecedent_gold_labels(pruned_gold_labels,
antecedent_labels)
# Now, compute the loss using the negative marginal log-likelihood.
# This is equal to the log of the sum of the probabilities of all antecedent predictions
# that would be consistent with the data, in the sense that we are minimising, for a
# given span, the negative marginal log likelihood of all antecedents which are in the
# same gold cluster as the span we are currently considering. Each span i predicts a
# single antecedent j, but there might be several prior mentions k in the same
# coreference cluster that would be valid antecedents. Our loss is the sum of the
# probability assigned to all valid antecedents. This is a valid objective for
# clustering as we don't mind which antecedent is predicted, so long as they are in
# the same coreference cluster.
coreference_log_probs = util.masked_log_softmax(coreference_scores, top_span_mask)
correct_antecedent_log_probs = coreference_log_probs + gold_antecedent_labels.log()
negative_marginal_log_likelihood = -util.logsumexp(correct_antecedent_log_probs).sum()
self._mention_recall(top_spans, metadata)
self._conll_coref_scores(top_spans, valid_antecedent_indices, predicted_antecedents, metadata)
output_dict["loss"] = negative_marginal_log_likelihood
if metadata is not None:
output_dict["document"] = [x["original_text"] for x in metadata]
return output_dict
def from_ability_score(cls, value):
if value < 1 or value > 30:
raise ValueError("Ability scores must be between 1 and 30 inclusive")
return cls(math.floor((value-10)/2))
ang_B/=360
A_time=0
B_time=0
Art=2*pi*r1/abs(s1)
Brt=2*pi*r2/abs(s2)
Af=2*pi*r1*ang_A/abs(s1)
Bf=2*pi*r2*ang_B/abs(s2)
print(Af,Bf)
Alis=list()
Blis=list()
if s1>0 and t>Af:
Alis.append(math.floor(Af))
A_time+=Af
if s2>0 and t>Bf:
Blis.append(math.floor(Bf))
B_time+=Bf
r=0
while(A_time<t or B_time<t):
if not r%2:
A_time+=Art-Af
B_time+=Brt-Bf
else:
A_time+=Af
B_time+=Bf
Alis.append(math.floor(A_time))
def print_total_average(total_average):
print('Total average: {:0.1f} (exact: {:0.5f})'.format(math.floor(total_average * 10) / 10, total_average))
import math as m print(m.floor(2.9)) print(m.pi) print(m.e) print(m.sqrt(25)) print(m.pow(2, 3))
def process(self, O, H, L, C):
self.grid_divider = self.box_size
self.boxes = []
last_box_level = 0 # последнее значение уровня
direction = 0 # направление
start_point_0 = 0.
start_point = 0.
box_size = self.box_size
boxes_to_reverse = self.boxes_to_reverse
data_length = H.shape[0]
self.levels = np.zeros(data_length)
opposite_boxes = 0.
for i in range(data_length):
if direction == 0: # начальная точка
if C[i] > O[i]:
direction = 1
#start_point = (floor(L[i] / self.grid_divider) + 1) * self.grid_divider
start_point = (floor(L[i] / self.grid_divider)) * self.grid_divider
#start_point = L[i]
extremum = H[i]
else:
direction = -1
start_point = (floor(H[i] / self.grid_divider)) * self.grid_divider
#start_point = H[i]
extremum = L[i]
#self.start_point_graph = start_point # - direction * 0 * box_size# + direction * box_size
distance_from_start = direction * (extremum - start_point) # positive value
boxes_from_start = int(floor(distance_from_start / box_size))
self.log('start {} dir {} boxes {}'.format(start_point, direction, boxes_from_start))
# if boxes_from_start < boxes_to_reverse:
# start_point = start_point - direction * (boxes_to_reverse - boxes_from_start) * box_size
# boxes_from_start = boxes_to_reverse
#self.start_point_graph = start_point - direction * box_size
self.start_point_graph = start_point
last_box_level = start_point + direction * boxes_from_start * box_size
boxes = direction * boxes_from_start
self.log(boxes)
self.boxes.append(boxes)
self.log('start {} dir {} boxes {}'.format(start_point, direction, boxes_from_start))
elif direction == 1 or direction == -1:
new_last_box_level = -1.
if direction == 1:
continue_level = H[i]
opposite_level = L[i]
sign = 1
else:
continue_level = L[i]
opposite_level = H[i]
sign = -1
if sign * (continue_level - last_box_level) >= box_size:
distance_from_start = (continue_level - start_point)
boxes_from_start = int(floor(distance_from_start / box_size))
new_last_box_level = start_point + boxes_from_start * box_size
#last_box_level = start_point + direction * boxes_from_start * box_size
opposite_distance = sign * (last_box_level - opposite_level)
opposite_boxes = int(floor(opposite_distance / box_size))
if opposite_boxes >= boxes_to_reverse and new_last_box_level < 0:
direction = - direction
self.boxes.append(direction * (opposite_boxes))
#self.boxes.append(direction * (opposite_boxes - 1))
#self.boxes.append(-(opposite_boxes - 2))
start_point = last_box_level
last_box_level = start_point + direction * opposite_boxes * box_size
#boxes_from_start = opposite_boxes
self.log('{} opposite boxes {} new level {} {}'.format(i, opposite_boxes, last_box_level, direction))
elif new_last_box_level > 0:
last_box_level = new_last_box_level
#self.boxes[-1] = boxes_from_start - direction
self.boxes[-1] = boxes_from_start
self.log('{} new boxes {} new level {} {}'.format(i, boxes_from_start, last_box_level, direction))
else:
raise Exception('Wrong direction {}!'.format(direction))
self.levels[i] = last_box_level
#print(direction, opposite_boxes, last_box_level, self.boxes[-1])
def interpolate(x,
scale=None,
output_size=None,
mode='linear',
align_corners=None):
'''
Resize an ND array with interpolation.
Scaling factors for spatial dimensions are determined by either
``scale`` or ``output_size``.
``nd = len(scale)`` or ``nd = len(output_size)`` determines the number of
spatial dimensions, and the last ``nd`` dimensions of the input ``x`` are
considered as the spatial dimensions to be resized.
If ``scale`` is given, the ``output_size`` is calculated by
``output_size[i] = floor(scale[i] * x.shape[i - len(scale)])``.
Example:
.. code-block:: python
import numpy as np
import nnabla as nn
import nnabla.functions as F
x_data = np.random.rand(64, 3, 224, 224)
x = nn.Variable.from_numpy_array(x_data)
# Resize by scales
y = F.interpolate(x, scale=(2, 2), mode='linear')
print(y.shape) # (64, 3, 448, 448)
y.forward()
print(y.d) # Print output
# Resize to a size
y2 = F.interpolate(x, output_size=(320, 257), mode='linear')
print(y2.shape) # (64, 3, 320, 257)
y2.forward()
print(y2.d) # Print output
Args:
x(~nnabla.Variable): N-D array with an arbitrary number of dimensions.
scale(tuple of ints): Scale factors along axes. The default is
``None``, and if this is omitted, ``output_size`` must be specified.
output_size(tuple of ints): The output sizes for axes. If this is
given, the scale factors are determined by the output sizes and the
input sizes. The default is ``None``, and if this is omitted,
``scale`` must be specified.
mode(str): Interpolation mode chosen from ('linear'|'nearest').
The default is 'linear'.
align_corners(bool): If true, the corner pixels of input and output
arrays are aligned, such that the output corner pixels have the
same values with the input corner pixels.
The default is ``None``, and it becomes ``True`` if mode is
'linear', otherwise ``False``.
Returns:
~nnabla.Variable: N-D array.
'''
from .function_bases import interpolate as interpolate_base
import math
if scale is None and output_size is None:
raise ValueError('Either scale or output_size must be given')
elif output_size is None:
output_size = [
int(math.floor(s * d))
for d, s in zip(x.shape[-len(scale):], scale)
]
if align_corners is None:
if mode == 'linear':
align_corners = True
else:
align_corners = False
return interpolate_base(x, output_size, mode, align_corners)
def make_circuit(n:int,f) -> QuantumCircuit:
# circuit begin
input_qubit = QuantumRegister(n,"qc")
classical = ClassicalRegister(n, "qm")
prog = QuantumCircuit(input_qubit, classical)
prog.h(input_qubit[0]) # number=3
prog.h(input_qubit[1]) # number=4
prog.h(input_qubit[0]) # number=57
prog.cz(input_qubit[4],input_qubit[0]) # number=58
prog.h(input_qubit[0]) # number=59
prog.cx(input_qubit[4],input_qubit[0]) # number=63
prog.z(input_qubit[4]) # number=64
prog.cx(input_qubit[4],input_qubit[0]) # number=65
prog.cx(input_qubit[4],input_qubit[0]) # number=56
prog.h(input_qubit[2]) # number=50
prog.cz(input_qubit[4],input_qubit[2]) # number=51
prog.h(input_qubit[2]) # number=52
prog.h(input_qubit[2]) # number=5
prog.h(input_qubit[3]) # number=6
prog.h(input_qubit[4]) # number=21
Zf = build_oracle(n, f)
repeat = floor(sqrt(2 ** n) * pi / 4)
for i in range(repeat):
prog.append(Zf.to_gate(), [input_qubit[i] for i in range(n)])
prog.h(input_qubit[0]) # number=1
prog.h(input_qubit[1]) # number=2
prog.h(input_qubit[2]) # number=7
prog.h(input_qubit[3]) # number=8
prog.h(input_qubit[0]) # number=28
prog.cx(input_qubit[3],input_qubit[0]) # number=60
prog.z(input_qubit[3]) # number=61
prog.cx(input_qubit[3],input_qubit[0]) # number=62
prog.cz(input_qubit[1],input_qubit[0]) # number=29
prog.h(input_qubit[0]) # number=30
prog.h(input_qubit[0]) # number=43
prog.cz(input_qubit[1],input_qubit[0]) # number=44
prog.h(input_qubit[0]) # number=45
prog.cx(input_qubit[1],input_qubit[0]) # number=35
prog.cx(input_qubit[1],input_qubit[0]) # number=38
prog.x(input_qubit[0]) # number=39
prog.cx(input_qubit[1],input_qubit[0]) # number=40
prog.cx(input_qubit[1],input_qubit[0]) # number=37
prog.h(input_qubit[0]) # number=46
prog.cz(input_qubit[1],input_qubit[0]) # number=47
prog.h(input_qubit[0]) # number=48
prog.cx(input_qubit[1],input_qubit[0]) # number=27
prog.x(input_qubit[1]) # number=10
prog.x(input_qubit[2]) # number=11
prog.x(input_qubit[3]) # number=12
if n>=2:
prog.mcu1(pi,input_qubit[1:],input_qubit[0])
prog.x(input_qubit[0]) # number=13
prog.cx(input_qubit[0],input_qubit[1]) # number=22
prog.y(input_qubit[2]) # number=41
prog.x(input_qubit[1]) # number=23
prog.cx(input_qubit[0],input_qubit[1]) # number=24
prog.rx(1.0398671683382215,input_qubit[2]) # number=31
prog.x(input_qubit[2]) # number=15
prog.x(input_qubit[3]) # number=16
prog.h(input_qubit[0]) # number=17
prog.h(input_qubit[1]) # number=18
prog.h(input_qubit[2]) # number=19
prog.h(input_qubit[3]) # number=20
# circuit end
return prog
def processCurrentImage(table):
imp = WindowManager.getCurrentImage()
fileName = imp.getTitle()
middleSlice = int(math.floor(imp.getNFrames() / 2.0) + (imp.getNFrames() % 2))
imp.setSlice(middleSlice)
IJ.run("Duplicate...", " ")
imp.close()
imp = WindowManager.getCurrentImage()
dir = fiji.analyze.directionality.Directionality_()
dir.setImagePlus(imp)
dir.setMethod(fiji.analyze.directionality.Directionality_.AnalysisMethod.FOURIER_COMPONENTS)
dir.setBinNumber(90)
dir.setBinStart(-90)
dir.setBuildOrientationMapFlag(False)
dir.computeHistograms()
dir.fitHistograms()
results = dir.getFitAnalysis()
direction = math.degrees(results[0][0])
dispersion = math.degrees(results[0 ][1])
amount = results[0][2]
goodness = results[0][3]
IJ.run("Clear Results")
IJ.run("FFT")
fftImp = WindowManager.getCurrentImage()
IJ.run("Mean...", "radius=2");
IJ.run("Find Maxima...", "noise=15 output=[Point Selection]")
IJ.run("Measure")
fftImp.changes = False
fftImp.close()
rt = ResultsTable.getResultsTable()
size = rt.size()
numberOfFrequences = size
if size>=5:
numberOfFrequences = 5
R = zeros('f', numberOfFrequences)
Theta = zeros('f', numberOfFrequences)
for i in range(0, numberOfFrequences):
R[i] = rt.getValue("R", i)
Theta[i] = rt.getValue("Theta", i)
table.incrementCounter()
table.addValue('image', fileName)
table.addValue('Direction', direction)
table.addValue('Dispersion', dispersion)
table.addValue('Amount', amount)
table.addValue('Goodness', goodness)
for i in range(0, numberOfFrequences):
table.addValue('R'+str(i), R[i])
table.addValue('Theta'+str(i), Theta[i])
widths = measureWidth(numberOfWidthMeasurements)
i = 1;
for width in widths:
table.addValue("width" + str(i), width)
i = i + 1
headings = rt.getHeadings()
for heading in headings:
if heading != "Label":
value = rt.getValue(heading, 0)
table.addValue(heading, value)
table.show('Directonality analysis')
def split_to_traintest(in_list,percentage):
n_train = math.floor(len(in_list)*percentage/100)
n_test = len(in_list) - n_train
return in_list[0:n_train],in_list[n_train:]
def _query_histograms(self):
measurements = []
for column_name_lower in self.scan_columns:
scan_column: ScanColumn = self.scan_columns[column_name_lower]
column_name = scan_column.column_name
if scan_column.is_metric_enabled(Metric.HISTOGRAM) and scan_column.numeric_expr:
buckets: int = scan_column.get_histogram_buckets()
min_value = scan_column.get_metric_value(Metric.MIN)
max_value = scan_column.get_metric_value(Metric.MAX)
if scan_column.has_numeric_values and min_value and max_value and min_value < max_value:
# Build the histogram query
min_value = floor(min_value * 1000) / 1000
max_value = ceil(max_value * 1000) / 1000
bucket_width = (max_value - min_value) / buckets
boundary = min_value
boundaries = [min_value]
for i in range(0, buckets):
boundary += bucket_width
boundaries.append(round(boundary, 3))
group_by_cte = scan_column.get_group_by_cte()
numeric_value_expr = scan_column.get_group_by_cte_numeric_value_expression()
field_clauses = []
for i in range(0, buckets):
lower_bound = '' if i == 0 else f'{boundaries[i]} <= {numeric_value_expr}'
upper_bound = '' if i == buckets - 1 else f'{numeric_value_expr} < {boundaries[i + 1]}'
optional_and = '' if lower_bound == '' or upper_bound == '' else ' and '
field_clauses.append(
f'SUM(CASE WHEN {lower_bound}{optional_and}{upper_bound} THEN frequency END)')
fields = ',\n '.join(field_clauses)
sql = (f'{group_by_cte} \n'
f'SELECT \n'
f' {fields} \n'
f'FROM group_by_value')
if self.filter_sql:
sql += f' \nWHERE {self.scan.filter_sql}'
row = self.warehouse.sql_fetchone(sql)
# Process the histogram query
frequencies = []
for i in range(0, buckets):
frequency = row[i]
frequencies.append(0 if not frequency else int(frequency))
histogram = {
'boundaries': boundaries,
'frequencies': frequencies
}
self._log_and_append_query_measurement(
measurements, Measurement(Metric.HISTOGRAM, column_name, histogram))
self._flush_measurements(measurements)
#author: moyuweiqing
#情感分析-by-snownlp&matplotlib
import os
from snownlp import SnowNLP
import matplotlib.pyplot as plt
import numpy as np
import math
path = os.path.abspath('..')
text = open(path + '\dependence\小王子.txt')
text = text.read()
s1 = text.replace('\n', '').replace(' ', '').replace('.', '。')#去除换行
#建立情感分析
sn1 = SnowNLP(s1)
sentimentslist = []
for i in sn1.sentences:
j = SnowNLP(i)
sentimentslist.append(j.sentiments)
#可视化处理，使用matplotlib
dic = {}
for i in np.arange(0, 1, 0.02):
index = round(i, 2)
dic[index] = 0
for i in sentimentslist:
temp = round(math.floor(i/0.02)*0.02, 2)
dic[temp] = dic[temp] + 1
plt.hist(sentimentslist,bins=np.arange(0,1,0.02))
plt.savefig(path+'\Results\sentimental_analysis（小王子）.png')
