ebenolson

"""
A deep neural network with or w/o dropout in one file.
"""

import numpy, theano, sys, math
from theano import tensor as T
from theano import shared
from theano.tensor.shared_randomstreams import RandomStreams
from collections import OrderedDict

ebenolson

"""
Functions to create network diagrams from a list of Layers.

Examples:

    Draw a minimal diagram to a pdf file:
        layers = lasagne.layers.get_all_layers(output_layer)
        draw_to_file(layers, 'network.pdf', output_shape=False)

    Draw a verbose diagram in an IPython notebook:

ebenolson

ebenolson

# Alec Radford, Indico, Kyle Kastner
# License: MIT
"""
Convolutional VAE in a single file.
Bringing in code from IndicoDataSolutions and Alec Radford (NewMu)
Additionally converted to use default conv2d interface instead of explicit cuDNN
"""
import theano
import theano.tensor as T
from theano.compat.python2x import OrderedDict

ebenolson

from lasagne.layers import Layer

class HighwayLayer(Layer):
    def __init__(self, incoming, layer_class, gate_nonlinearity=None,
                 **kwargs):
        super(HighwayLayer, self).__init__(incoming)

        self.H_layer = layer_class(incoming, **kwargs)
        
        if gate_nonlinearity:

ebenolson

# Usage: experiment.sh SOURCE_FILE LOG_COMMIT_MESSAGE

if [ $# -eq 0 ]
    then
        echo "Usage: experiment.sh SOURCE_FILE [LOG_COMMIT_MESSAGE]"
        exit -1
fi

if [ $# -eq 2 ]
    then

ebenolson

require 'optim'
cj = require('cjson')

a = torch.Tensor({0.1,0.2,0.3})
x0 = torch.Tensor({1, 1, 1})

function f(x) 
    return torch.sum(torch.cmul(a, torch.pow(x, 2)))
end

ebenolson

name: "CIFAR10_full"
input: "data"
input_shape {
  dim: 1
  dim: 3
  dim: 32
  dim: 32
}
layers {
  name: "conv1"

ebenolson

import numpy as np
import gc
import memory_profiler

import theano
import theano.tensor as T
from theano.sandbox.cuda.dnn import dnn_conv

X = T.tensor4()
W = T.tensor4()

ebenolson

from lasagne.layers import InputLayer
from lasagne.layers import DenseLayer
from lasagne.layers import ConcatLayer
from lasagne.layers import NonlinearityLayer
from lasagne.layers import GlobalPoolLayer
from lasagne.layers.dnn import Conv2DDNNLayer as ConvLayer
from lasagne.layers.dnn import MaxPool2DDNNLayer as PoolLayerDNN
from lasagne.layers import MaxPool2DLayer as PoolLayer
from lasagne.layers import LocalResponseNormalization2DLayer as LRNLayer
from lasagne.nonlinearities import softmax, linear