def model(weights, features):
'''Calculate regression for features given weights.
[number] [number] --> number
'''
return stats.dotprod(weights, features)
del data
# invert negative examples; put them in positive set
for d in neg:
pos.append([-f for f in d])
assert len(pos) == count
del neg
# initialize the weight vector
weights = len(pos) * [0.0]
# iterate
iteration = 0
mistakes = 1
while mistakes: # should be while iteration < (m * |v|^2) / (d^2)
iteration += 1
mistakes = 0
for d in pos:
if stats.dotprod(weights, d) > 0:
pass # good!
else:
weights = [w + f for w, f in zip(weights, d)]
mistakes += 1
print 'Iteration {}, total mistakes {}'.format(iteration, mistakes)
print 'Classifier weights: {}'.format(
' '.join([str(w) for w in weights]))
print 'Normalized with threshold: {}'.format(
' '.join([str(w/weights[0]) for w in weights[1:]]))
