machine_learning_algos.py

#!/usr/bin/env python3

from statistics import mean, stdev

def squared_error_cost(hypothesis, data):
    return sum( (hypothesis(point[:-1]) - point[-1])**2 for point in data ) / 2*len(data)

def linear_optimize_gradient_descent(data, args, alpha=.01, max_iters=1000):
    """Takes a sequence of n-tuples, where n-1 is the number of features of the model, and the last index is the output"""
    data = [(1,) + tuple(point) for point in data]
    feature_scale_parameters = [(mean(feature), stdev(feature)) for feature in zip(*data)]
    data = [tuple((x - m)/s for x,(m,s) in zip(point, feature_scale_parameters)) for point in data]
    next_args = list(args)
    for i in range(max_iters):
        hyp = create_linear_hypothesis(args)
        for feature in range(len(data[0])):
            next_args[feature] -= alpha/len(data) * sum((hyp(point) - point[1])*point[feature] for point in data)
        args = list(next_args)
    args = [arg*s + m for arg, (m,s) in zip(args, feature_scaling_parameters)]
    return args

def create_linear_hypothesis(args):
    return lambda vals: sum(arg*val for (arg,val) in zip(args, (1,) + tuple(vals)))