a5_brute_force_training_partial.py

from copy import deepcopy

def train(X, y, model_parameters, step=0.1, iterations=100): 
    # Make prediction for every data sample
    predictions = [predict(x, model_parameters) for x in X]

    # Calculate initial cost for model - MSE
    lowest_error = mse(predictions, y)

    for i in range(iterations):
        candidates, errors = list(), list()

        # w increased, b increased
        param_candidate = deepcopy(model_parameters)
        param_candidate["b"] += step
        param_candidate["w"] += step
        
        candidate_pred = [predict(x, param_candidate) for x in X]
        candidate_error = mse(candidate_pred, y)

        candidates.append(param_candidate)
        errors.append(candidate_error)

        # w increased, b unchanged
        (...)

        # w increased, b decreased
        (...)

        # w unchanged, b increased
        (...)

        # w unchanged, b unchanged
        (...)

        # w unchanged, b decreased
        (...)
        
        # w decreased, b increased
        (...)

        # w decreased, b unchanged
        (...)

        # w decreased, b decreased
        (...)

        # Update with parameters for which loss is smallest
        best_candidate = None
        for candidate, candidate_error in zip(candidates, errors):
            if candidate_error < lowest_error:
                lowest_error = candidate_error
                model_parameters["w"] = candidate["w"]
                model_parameters["b"] = candidate["b"]