acyclic affine allocator arccosine backend backprop backpropagation Bernoulli bilinear bool boolean broadcast broadcasted Caffe Chainer Cifer channelwise compatibilities composability composable compositional connectionist concat contrastive convolutional covariance CPU CTC cuBLAS CUDA cuDNN CuPy cuRAND customizable cutorch dataset datasets de deallocated deallocation deconvolution deconvolutional deserialization deserialize deserialized deserializes deserializer deserializers deserializing differentiable dimensionalities dimensionality dimentional dtype dtypes elementwise embeddings evaluator facto FFT finalizer functionalities gaussian GPU grey greyscale gzip hdf huber hyperparameter hyperparameters hypoteneous implementers indices infeasible initializer initializers instantiation iterable iteratively libhdf learnable Linux maxout memoization memoize memoized memoizes memoizing minibatch mlpconv mnist multi multiclass namespace NaN normalizer npz NumPy online optimizer optimizers outliers Overfeat parallelization parallelizations parallelize parallelized parallelizes parallelizing parameterize parameterized perceptron Perf permutate permutates pluggable postprocess pre preprocessing profiler profilers proto PRs quantization reimplement reinstall resized runtime serializable serializer serializers sharding siamese sigmoid softmax stateful subdirectories subdirectory summarization th Theano treebank tuple tuples ufunc ufuncs unigram uninstall unnormalize unnormalized unpooling unregisters unuseful variadic variational vecLib vectorized versioning workspace Bengio Cho Courville farley Goodfellow Jurgen Mirza Nesterov Schmidhuber Socher Warde Zeiler