def logmode(points,K=5,burst=None):
'''
Accepts list of ISI times.
Finds the mode using a log-KDE density estimate
'''
points = np.array(points)
if not burst is None:
points = points[points>burst] # remove burst
x,y = kdepeak(np.log(K+points[points>0]))
x = np.exp(x)-K
y = y/(K+x)
mode = x[np.argmax(y)]
return mode
def modefind(points,burst=10):
'''
Removes intervals shorter than 10m
Finds peak using log-KDE approximation
'''
points = np.array(points)
points = points[points>burst] # remove burst
K = 5
x,y = kdepeak(np.log(K+points[points>0]))
x = np.exp(x)-K
y = y/(K+x)
mode = x[np.argmax(y)]
return mode
def peakfinder5(st,K=5):
'''
Found this with the old unit classification code.
Haven't had time to reach it and check out what it does
'''
points = np.diff(st)
points = np.array(points)
points = points[points>10] # remove burst
n, bins, patches = hist(points,
bins=np.linspace(0,500,251),
facecolor='k',
normed=1)
centers = (bins[1:]+bins[:-1])/2
x,y = kdepeak(points,
x_grid=np.linspace(0,500,251))
plot(x,y,color='r',lw=1)
p1 = x[np.argmax(y)]
x,y = kdepeak(np.log(K+points[points>0]))
x = np.exp(x)-K
y = y/(K+x)
plt.plot(x,y,color='g',lw=1)
p2 = x[np.argmax(y)]
plt.xlim(0,500)
return p1,p2
def logmodeplot(points,K=5,burst=None):
'''
Accepts list of ISI times.
Finds the mode using a log-KDE density estimate
Plots this along with histogram
'''
points = np.array(points)
if not burst is None:
points = points[points>burst] # remove burst
x,y = kdepeak(np.log(K+points[points>0]))
x = np.exp(x)-K
y = y/(K+x)
cla()
plt.hist(points,60,normed=1,color='k')
plt.plot(x,y,lw=2,color='r')
ybar(x[np.argmax(y)],color='r',lw=2)
plt.draw()
plt.show()
mi = np.argmax(y)
mode = x[mi]
return mode
