def compare_mds_tsne(dataset='mnist', perplexity=30):
data = samples.load(dataset)
D = [data['X']] * 2
va = {'perplexity': perplexity}
mv = MPSE(D,
visualization_method=['mds', 'tsne'],
visualization_args=va,
colors=data['colors'],
verbose=2)
mv.gd()
mv.plot_computations()
mv.plot_embedding(title='final embeding')
mv.plot_images()
plt.draw()
plt.pause(0.2)
plt.show()
return
def plot(p1, p2, name):
gsamps = samples.load('gsa_samples/GSA_Samples.data')
good_samps = [sample for sample in gsamps if sample.pb is not None]
pb_vals = [fix(sample.pb) for sample in good_samps]
cs_vals = [sample.cs for sample in good_samps]
cult_vals = [sample.cultivated for sample in good_samps]
art_vals = [sample.artificial for sample in good_samps]
rain_vals = [sample.rain.mean for sample in good_samps]
relief_vals = [sample.relief.mean for sample in good_samps]
elevation_vals = [sample.elevation.mean for sample in good_samps]
slope_vals = [sample.slope.mean for sample in good_samps]
area_vals = [sample.area/1000000 for sample in good_samps]
pb = (pb_vals, LABELS['pb'], 'pb')
cs = (cs_vals, LABELS['cs'], 'cs')
cult = (cult_vals, LABELS['cult'], 'cult')
art = (art_vals, LABELS['art'], 'art')
rain = (rain_vals, LABELS['rain'], 'rain')
relief = (relief_vals, LABELS['relief'], 'relief')
elevation = (elevation_vals, LABELS['elevation'], 'elev')
slope = (slope_vals, LABELS['slope'], 'slope')
area = (area_vals, LABELS['area'], 'area')
mapper = {
'pb':pb,
'cs':cs,
'cult':cult,
'cultivated':cult,
'cultivation':cult,
'art':art,
'artificial':art,
'rain':rain,
'precip':rain,
'precipitation':rain,
'relief':relief,
'elevation':elevation,
'elev':elevation,
'slope':slope,
'area':area,
}
oname = "gsa_samples/plots/%s" % name
samples.plot(mapper[p1], mapper[p2], oname)
def gogographit(neg='nothing'):
gsamps = samples.load('gsa_samples/GSA_Samples.data')
good_samps = [sample for sample in gsamps if sample.pb is not None]
if neg == 'remove':
good_samps = [sample for sample in good_samps if sample.pb >= 0]
if neg == 'zero':
pb_vals = [fix(sample.pb) for sample in good_samps]
else:
pb_vals = [sample.pb for sample in good_samps]
cs_vals = [sample.cs for sample in good_samps]
cult_vals = [sample.cultivated for sample in good_samps]
art_vals = [sample.artificial for sample in good_samps]
rain_vals = [sample.rain.mean for sample in good_samps]
relief_vals = [sample.relief.mean for sample in good_samps]
elevation_vals = [sample.elevation.mean for sample in good_samps]
slope_vals = [sample.slope.mean for sample in good_samps]
area_vals = [sample.area/1000000 for sample in good_samps]
pb = (pb_vals, LABELS['pb'], 'pb')
cs = (cs_vals, LABELS['cs'], 'cs')
cult = (cult_vals, LABELS['cult'], 'cult')
art = (art_vals, LABELS['art'], 'art')
rain = (rain_vals, LABELS['rain'], 'rain')
relief = (relief_vals, LABELS['relief'], 'relief')
elevation = (elevation_vals, LABELS['elevation'], 'elev')
slope = (slope_vals, LABELS['slope'], 'slope')
area = (area_vals, LABELS['area'], 'area')
indvars = (cult, art, rain, relief, elevation, slope, area)
plot_folder = 'gsa_samples/plots'
if neg == 'zero':
extra = "_zero"
elif neg == 'remove':
extra = "no-neg"
else:
extra = ""
for var in indvars:
print "plotting %s against %s" % (var[2], 'pb')
fname = "%s/%s_pb%s" % (plot_folder, var[2], extra)
samples.plot(var, pb, fname)
#set default values to known parameters
total_field_size = cf.SIZE_X*cf.SIZE_Y
layer_sizes = [total_field_size*3, total_field_size*2, total_field_size,cf.SIZE_X]
if args.layer_sizes:
print("customn layer size")
layer_sizes = args.layer_sizes
#create AI
ai = neural.NeuralAI(layer_sizes=layer_sizes, learning_rate=args.learning_rate, momentum=args.momentum)
#ai = ai.read_model_data("best_ai")
if args.ai_file:
print("reading")
ai.read_model_data(args.ai_file)
else:
td = samples.load(args.training_file)
print("learning")
ai.learn_epoch(td,args.epochs)
def shuffle():
print("shuffling")
random.shuffle(td)
def learn():
print("learning")
ai.learn(td)
def save():
ai.write_model_data("ai")
def load(tgt):
return cv2.resize(samples.load(int(tgt)), (WIDTH, HEIGHT))
def load_samples():
df = samples.load()
return df.loc['chronic']
