Lasagne2.py

# -*- coding: utf-8 -*-
"""
Created on Mon Aug 31 15:03:52 2015

@author: tmold_000
"""
import numpy as np
import Normalizers
from lasagne import layers
from lasagne.updates import nesterov_momentum
from nolearn.lasagne import NeuralNet
import lasagne
import dataParser
import matplotlib.pyplot as plt
import time
import pickle

numInputs = 31;
numOutputs = 28;
def trainNet(X, Y, ln, loadFile = ""):
    net1 = NeuralNet(
        layers=[  # four layers: two hidden layers
            ('input', layers.InputLayer),
            ('hidden', layers.DenseLayer),
            ('hidden1', layers.DenseLayer),
            ('output', layers.DenseLayer),
            ],
        # layer parameters: Best 400 400
        input_shape=(None, numInputs),  # 31 inputs
        hidden_num_units=400,  # number of units in hidden layer
        hidden1_num_units=400,
        hidden_nonlinearity=lasagne.nonlinearities.sigmoid,
        hidden1_nonlinearity=lasagne.nonlinearities.sigmoid,
        output_nonlinearity=None,  # output layer uses identity function
        output_num_units=numOutputs,  # 4 outputs
    
        # optimization method:
        update=nesterov_momentum,
        update_learning_rate=ln,
        update_momentum=0.9,
    
        regression=True,  # flag to indicate we're dealing with regression problem
        max_epochs=1500,  # we want to train this many epochs
        verbose=1,
        )
    #if (loadFile != ""):
        #net1.load_params_from(loadFile)
    net1.max_epochs = 10
    net1.update_learning_rate = ln;
    net1.fit(X, Y) # This thing try to do the fit itself
    return net1


    
#generations, generationsToInputs, generationsToOutputs = dataParser.parse(fname = "whole_population_0.txt")
iters = 1
saveFile = "LasagneWeights400_2Layer"
trainingInputs, trainingOutputs, testInputs, testOutputs = dataParser.makeSets(generationsToInputs, generationsToOutputs, generations[0:200], 0.5, 0.25)
ln = 0.01
for n in range(iters): 
    X = Normalizers.gaussNormalize(trainingInputs)
    Xtest = Normalizers.gaussNormalize(testInputs)
    #
    Y = Normalizers.gaussNormalize(trainingOutputs)
    Ytest = Normalizers.gaussNormalize(testOutputs)
    #
    #X = trainingInputs
    #Xtest = testInputs
    #Y = trainingOutputs
    #Ytest = testOutputs
    
    net = trainNet(np.asarray(X,np.float32), np.asarray(Y,np.float32), ln, saveFile)
    y_pred = net.predict(np.asarray(Xtest,np.float32))
    errors = Normalizers.deGauss(trainingOutputs, Ytest) - Normalizers.deGauss(trainingOutputs, y_pred)
    print(np.mean(abs(errors),axis = 0));
    avgError = np.mean(np.sum(abs(errors), axis = 1))
    print(avgError) #Show the average total error
    plt.hist(Normalizers.reject_outliers(np.sum(abs(errors), axis = 1)),20)
    # Saving the objects:
with open('model.pickle', 'w') as f:
    pickle.dump(net, f)
    plt.show()
    plt.title(avgError)
    #plt.savefig("Two400")
    #net.save_params_to(saveFile)