def solve(self, pos):
self.pos = pos
ret1 = OvR.separate(self.matx, self.label, pos, self.m, self.n)
mat, vec = Smote.genNew(ret1[0], ret1[1], self.n)
# mat, vec = BorderlineSmote.genNew(ret1[0], ret1[1], self.n)
beta = Descent.solve(mat, vec)
'''
ensembleN = 5
beta = np.zeros((self.n))
for i in range(ensembleN):
mat, vec = EasyEnsemble.genNew(ret1[0], ret1[1], self.n)
beta = beta + Descent.solve(mat, vec)
beta = beta / ensembleN
'''
print('beta', beta)
return beta
for i in range(3, 10):
# read from file
with open('../data/Data/0' + str(i) + '/watch/raw_4_0_unsync.txt') as f:
x = f.readlines()
with open('../data/Data/10/watch/raw_1_0_unsync.txt') as f:
x = f.readlines()
data = []
for e in x:
data.append(map(float, e[:-1].split(' ')[1:-1]))
#print len(data), len(data[0]), data[0]
# --- get a gesture chunk ---
d_obj = Descent()
g_obj = WatchGesture()
swipe_threshold = int(d_obj.median(data[0])) * 0.80
touch_threshold = int(d_obj.median(data[0])) * 0.10
#swipe_threshold, touch_threshold = d_obj.auto_thresh(data)
y = g_obj.gd(data, swipe_threshold, 1)
if len(y) > 3:
print '~~~~\nSwipe Detected, Swipe gesture direction\n'
print len(y), y
class Gesture(object):
def __init__(self):
self.m = 0
self.y = []
self.obj = Descent()
def clearobj(self):
self.m = 0
self.y = []
def gd(self, data, threshold, control=5):
i = 0
temp = -1
y = []
while i < len(data) - 1:
temp = self.obj.gradient_descent(data[i], threshold)
while temp == -1 and i < len(data) - 1:
i += 1
temp = self.obj.gradient_descent(data[i], threshold)
w = []
temp = self.obj.gradient_descent(data[i], threshold)
while temp != -1 and i < len(data) - 1:
w.append(temp)
i += 1
temp = self.obj.gradient_descent(data[i], threshold)
# now w has one chunk
cnt = 0
for j in range(1, len(w)):
if w[j] > w[j - 1]:
cnt += 1
elif w[j] < w[j - 1]:
cnt -= 1
#print i, cnt
if cnt > control:
y.append(2)
elif cnt < -control:
y.append(1)
return y
def gd_double_touch(self, data, threshold, control=0):
i = 0
temp1 = -1
temp2 = -1
y = []
last_i = -1
while i < len(data) - 1:
temp1 = self.obj.gradient_descent(data[i], threshold)
temp2 = self.obj.gradient_descent(data[i][::-1], threshold)
while temp1 == -1 and i < len(data) - 1:
i += 1
temp1 = self.obj.gradient_descent(data[i], threshold)
temp2 = self.obj.gradient_descent(data[i][::-1], threshold)
w1 = []
w2 = []
temp1 = self.obj.gradient_descent(data[i], threshold)
temp2 = self.obj.gradient_descent(data[i][::-1], threshold)
while temp1 != -1 and i < len(data) - 1:
w1.append(temp1)
w2.append(15 - temp2)
i += 1
temp1 = self.obj.gradient_descent(data[i], threshold)
temp2 = self.obj.gradient_descent(data[i][::-1], threshold)
# now w has one chunk
print len(w1), len(w2), w1, w2
if len(w1) > 2:
if w1 == w2:
if i - last_i < 20:
y.pop()
y.append(
'd_' +
str(max(k
for k, v in Counter(w1).items() if v > 1)))
else:
y.append(max(w1))
else:
if y and i - last_i < 20:
y.pop()
y.append(
'm_' +
str(max(k for k, v in Counter(w1).items() if v > 1)) +
'_' +
str(max(k for k, v in Counter(w2).items() if v > 1)))
last_i = i
return y
def __init__(self):
self.m = 0
self.y = []
self.obj = Descent()
class WatchGesture(object):
def __init__(self):
self.m = 0
self.y = []
self.obj = Descent()
def clearobj(self):
self.m = 0
self.y = []
def findLongestConseqSubseq(self, arr):
# Time Complexity O(n) - Linear
s = set([])
ans = 0
sp = -1
# Hash all the array elements
for ele in arr:
s.add(ele)
# check each possible sequence from the start
# then update optimal length
for i in range(len(arr)):
# if current element is the starting
# element of a sequence
if (arr[i] - 1) not in s:
# Then check for next elements in the
# sequence
j = arr[i]
while (j in s):
j += 1
# update optimal length if this length
# is more
if j - arr[i] > ans:
ans = j - arr[i]
sp = arr[i]
#ans=max(ans, j-arr[i])
return ans, sp
def gd(self, data, threshold, control=5):
i = 0
temp = -1
y = []
#print len(data), len(data[0]), data[0]
tdata = []
for e in data:
tdata.append([e[:6], e[6:11], e[11:17], e[17:]])
while i < len(data) - 1:
temp = self.obj.gradient_descent(data[i], threshold)
while temp == -1 and i < len(data) - 1:
i += 1
temp = self.obj.gradient_descent(data[i], threshold)
w = []
temp = self.obj.gradient_descent(data[i], threshold)
while temp != -1 and i < len(data) - 1:
w.append(temp)
i += 1
temp = self.obj.gradient_descent(data[i], threshold)
# now w has one chunk
#print len(w), self.findLongestConseqSubseq(w)
cnt = 0
for j in range(1, len(w)):
if w[j] > w[j - 1]:
cnt += 1
elif w[j] < w[j - 1]:
cnt -= 1
if len(w) > 3:
a, b = self.findLongestConseqSubseq(w)
if b + a / 2 in range(0, 6) or b + a / 2 in range(11, 17):
if cnt > control:
y.append(2)
elif cnt < -control:
y.append(1)
#y.append('1/2')
elif b + a / 2 in range(6, 11) or b + a / 2 in range(17, 22):
if cnt > control:
y.append(4)
elif cnt < -control:
y.append(3)
return y