Cacher is the code snippet organizer for pro developers

#coding: utf-8
from autoencoder import *

def unpickle(f):
    fo = open(f, 'rb')
    d = cPickle.load(fo)
    fo.close()
    return d

def load_data(dataset):
    """データセットをロードしてGPUの共有変数に格納"""
    d = unpickle(dataset)
    data_x = np.array(d['data']) / 255.0

    def shared_dataset(data_x, borrow=True):
        shared_x = theano.shared(np.asarray(data_x, dtype=theano.config.floatX), borrow=borrow)
        return shared_x

    train_set_x = shared_dataset(data_x)

    return train_set_x


if __name__ == "__main__":
    learning_rate = 0.01
    training_epochs = 100
    batch_size = 20

    # 学習データのロード
    # 今回は評価はしないため訓練データのみ
    train_set_x = load_data('../data/cifar10/data_batch_1')
    print train_set_x.get_value().shape

    # ミニバッチ数
    n_train_batches = train_set_x.get_value(borrow=True).shape[0] / batch_size

    # ミニバッチのインデックスを表すシンボル
    index = T.lscalar()

    # ミニバッチの学習データを表すシンボル
    x = T.matrix('x')

    # モデル構築
    rng = np.random.RandomState(123)
    theano_rng = RandomStreams(rng.randint(2 ** 30))

    autoencoder = Autoencoder(numpy_rng=rng,
                               theano_rng=theano_rng,
                               input=x,
                               n_visible=3072,
                               n_hidden=100)

    # コスト関数と更新式のシンボルを取得
    cost, updates = autoencoder.get_cost_updates(learning_rate=learning_rate)

    # 訓練用の関数を定義
    train_da = theano.function([index],
        cost,
        updates=updates,
        givens={
            x: train_set_x[index * batch_size: (index + 1) * batch_size]
        })

    # モデル訓練
    fp = open("cost.txt", "w")
    start_time = time.clock()
    for epoch in xrange(training_epochs):
        c = []
        for batch_index in xrange(n_train_batches):
            c.append(train_da(batch_index))

        print "Training epoch %d, cost %f" % (epoch, np.mean(c))
        fp.write("%d\t%f\n" % (epoch, np.mean(c)))
        fp.flush()

    end_time = time.clock()
    training_time = (end_time - start_time)
    fp.close()

    print "time: %.2fm" % ((training_time / 60.0))

    # 学習したモデルの状態を保存
    f = open('autoencoder_cifar10.pkl', 'wb')
    cPickle.dump(autoencoder.__getstate__(), f, protocol=cPickle.HIGHEST_PROTOCOL)
    f.close()

    # 学習された重みを可視化
    import matplotlib.pyplot as plt
    W = autoencoder.W.get_value().T
    print W.shape

    W = (W - np.min(W)) / (np.max(W) - np.min(W))
    W *= 255.0
    W = W.astype(np.int)

    pos = 1
    for i in range(100):
        plt.subplot(10, 10, pos)
        plt.subplots_adjust(wspace=0, hspace=0)
        plt.imshow(W[i].reshape(3, 32, 32).transpose(1, 2, 0), interpolation='nearest')
        plt.axis('off')
        pos += 1
    plt.savefig("autoencoder_filters_cifar10.png")