#!/usr/bin/env python
# encoding: utf-8
import datetime
import numpy as np
import matplotlib.pyplot as pl
from sklearn.linear_model import LogisticRegression
from mpl_toolkits.mplot3d import Axes3D
from example.hallem.data import Hallem
import toydata
from _core import validation
hallem = Hallem()
data = np.transpose(hallem.response)
print data.shape
big_matrix, labels = toydata.generate_noise_samples(data)
c_range = np.arange(0.001, 0.008, 0.00001)
f = []
a = []
b1 = True
b2 = True
print "### regularizationaraitining"
## Probing over C
for c in c_range:
lr = LogisticRegression(penalty='l1', C=c)
#!/usr/bin/env python
# encoding: utf-8
"""
Compute the correlation one-to-all of a specified odorant.
"""
import scipy as sp
import pylab as pl
import numpy as np
from scipy import stats
from example.hallem.data import Hallem
hallem = Hallem()
data = hallem.response
o1 = hallem.get_odorant_index("E2 Hexenol")
o1_name = hallem.odorant_list[o1]
o1_spec = data[o1]
correlations = []
for i, spec in enumerate(data):
o2 = i
o2_name = hallem.odorant_list[o2]
o2_spec = spec
c = sp.stats.pearsonr(o1_spec, o2_spec)[0]
correlations.append(c)
print o2_name, c
correlations = np.round(np.asarray(correlations), 3)
"""
Analysis of stability of backward prediction.
Since we have no test set, like in many machine learning tasks, we generate
one from the original data.
For each glomerulus we create 100 samples and classify them according to the
glomerulus they are based on.
For different levels of noise, the accuracy of prediction is measured,
e.g. how many labels could be predicted correctly.
"""
import numpy as np
import matplotlib.pyplot as pl
from example.hallem.data import Hallem
from _core import featureselection, validation
data = np.transpose(Hallem().response)
# compute features
feature_list, scores = featureselection.backward_elimination(data)
# creating a list of feature lists with increasing size
top = []
for i in range(15):
top.append((feature_list[:i + 1])[::-1])
# # levels of noise which will be added
sd_range = range(0, 50, 10)
results = validation.validate(data, top, noise=sd_range)
np.set_printoptions(suppress=True, precision=3)
#!/usr/bin/env python
# encoding: utf-8
'''
Compute the euclidean distance of the reference set (deBruyne) on Hallem and
DoOR.
'''
import numpy as np
from scipy.spatial import distance
from example.hallem.data import Hallem
from example.dorsal_door import DoOR
hallem = Hallem()
data, or_list, odorants = hallem.get_reference_set()
title = "Diagnostic set from DeBruyne 2001 on Hallem"
path = "figures/debruyne_hallem.png"
#plotting.plot_fingerprints(title, odorants, data, or_list, path)
print "Euclidean distance Hallem:", np.min(distance.pdist(data, 'euclidean'))
door = DoOR()
door_data, door_ors, door_odorants = door.get_dorsal_reference_data()
print "Euclidean distance DoOR:", np.min(
distance.pdist(door_data, 'euclidean'))
title = "Diagnostic set from DeBruyne 2001 on dorsal DoOR"
path = "figures/debruyne_door.png"
#plotting.plot_fingerprints(title, door_odorants, door_data, door_ors, path)