Index
A
- AdaDelta / Optimization and other update rules
- Adagrad / Optimization and other update rules
- Adam / Optimization and other update rules
- AlphaGo / Q-learning
- analogical reasoning / Evaluating embeddings – analogical reasoning
- architecture
- designing, for model / Designing the architecture for the model
- vector representations, of words / Vector representations of words
- sentence representation, using bi-LSTM / Sentence representation using bi-LSTM
- outputting probabilities, with softmax classifier / Outputting probabilities with the softmax classifier
- artificial intelligence
- Association for Computational Linguistics (ACL) / Seq2seq for translation
- asynchronous gradient descent / Training stability
- attention mechanism
- differentiable / Differentiable mechanism of attention
- about / Differentiable mechanism of attention
- translations / Better translations with attention mechanism
- annotate images / Better annotate images with attention mechanism
- auto encoder / Deep belief bets
- automatic differentiation / Functions and automatic differentiation
B
- backpropagation / Backpropagation and stochastic gradient descent
- Backpropagation Through Time (BPTT) / Simple recurrent network
- Basic Linear Algebra Subprograms (BLAS) / Theano Op in Python for the GPU
- batch normalization / Batch normalization
- batch normalization layer / Batch normalization
- Bayes Network theory / Dropout for RNN
- beam search algorithm / Improving efficiency of sequence-to-sequence network
- broadcasting / Elementwise operators
C
- Character Error Rate (CER) / Metrics for natural language performance
- coalesced transpose
- via shared memory / Coalesced transpose via shared memory, NVIDIA parallel for all
- via NVIDIA parallel / Coalesced transpose via shared memory, NVIDIA parallel for all
- model conversions / Model conversions
- Conditional Random Fields (CRF) / Deconvolutions for images
- Continuous Bag of Words (CBOW) / Encoding and embedding
- continuous Bag of Words model / Continuous Bag of Words model
- controller / Store and retrieve information in Neural Turing Machines
- Convolutional Neural Network (CNN) / Encoding and embedding
- convolutions / Convolutions and max layers
- cost function / Cost function and errors
- CUDA
- URL, for downloading / GPU drivers and libraries
D
- data augmentation / Data augmentation
- dataset / Dataset
- for natural language / A dataset for natural language
- character level / A dataset for natural language
- word level / A dataset for natural language
- DeConvNet / Deconvolutions for images
- deep belief bets / Deep belief bets
- deep belief network (DBN) / Deep belief bets
- Deeplearning.net Theano
- references / Related articles
- DeepMask network / Deconvolutions for images
- DeepMind algorithm / Q-learning
- Deep Q-network / Deep Q-network
- deep transition network / Deep approaches for RNN
- deep transition recurrent network / Deep transition recurrent network
- dense connections / Dense connections
- dimension manipulation operators / Dimension manipulation operators
- dropout / Dropout
E
- elementwise operators / Elementwise operators
- embedding / Encoding and embedding
- encoding / Encoding and embedding
- episodic memory
- with dynamic memory networks / Episodic memory with dynamic memory networks
- errors / Cost function and errors
- external memory bank / Store and retrieve information in Neural Turing Machines
F
- functions / Functions and automatic differentiation
G
- gated recurrent network / Gated recurrent network
- GEneral Matrix to Matrix Multiplication (GEMM) / Theano Op in Python for the GPU
- General Matrix to Vector Multiplication (GEMV) / Theano Op in Python for the GPU
- generative adversarial networks (GANs)
- about / Generative adversarial networks
- improving / Improve GANs
- generative model
- about / Generative models
- Restricted Boltzmann Machine / Restricted Boltzmann Machines
- deep belief bets / Deep belief bets
- generative adversarial networks (GANs) / Generative adversarial networks
- global average pooling / Global average pooling
- graphs / Graphs and symbolic computing
- greedy approach / Deep Q-network
H
- highway networks design principle / Highway networks design principle
- host code / Theano Op in C for GPU
I
- identity connection / Residual connections
- identity connections / Highway networks design principle
- images
- deconvolutions / Deconvolutions for images
- Inceptionism / Deconvolutions for images
- Independent Component Analysis (ICA) / Visualizing the learned embeddings
- inference / Inference
- internal covariate shift / Batch normalization
- Intersection over Union (IOU) / Region-based localization networks
K
- Keras
- installing / Installing and configuring Keras
- configuring / Installing and configuring Keras
- programming / Programming with Keras
- SemEval 2013 dataset / SemEval 2013 dataset
- model, training / Compiling and training the model
- model, compiling / Compiling and training the model
- kernel / Theano Op in C for GPU
L
- Lasagne
- MNIST CNN model / MNIST CNN model with Lasagne
- Latent Sementic Analysis / Indexing (LSA / LSI) / Visualizing the learned embeddings
- layer input normalization / Batch normalization
- learned embeddings
- visualizing / Visualizing the learned embeddings
- linear algebra operators / Linear algebra operators
- Linear Discriminant Analysis (LDA) / Visualizing the learned embeddings
- localization network
- about / A localization network
- recurrent neural net, applied to images / Recurrent neural net applied to images
- Locally Linear Embedding (LLE) / Visualizing the learned embeddings
- Long Short-Term Memory (LSTM) / Sentence representation using bi-LSTM
- loops
- in symbolic computing / Loops in symbolic computing
- loss comparison
- training / Training loss comparison
- loss function
- classification / Classification loss function
- LSTM network / LSTM network
M
- max layers / Convolutions and max layers
- memory / Memory and variables
- memory networks
- about / Memory networks
- episodic memory, with dynamic memory networks / Episodic memory with dynamic memory networks
- MNIST CNN model
- with Lasagne / MNIST CNN model with Lasagne
- MNIST dataset / The MNIST dataset
- model
- training / Training the model
- compiling, in Keras / Compiling and training the model
- training, in Keras / Compiling and training the model
- evaluating / Evaluating the model
- loading / Saving and loading the model
- saving / Saving and loading the model
- example, executing / Running the example
- model collapse / Improve GANs
- Modified National Institute of Standards and Technology (MNIST) / The MNIST dataset
- momentum / Optimization and other update rules
- Monte Carlo Tree Search (MCTS) / Q-learning
- multi-GPU / Multi-GPU
- multi-layer perceptron (MLP) / Multiple layer model
- Multi Dimensional Scaling (MDS) / Visualizing the learned embeddings
- multimodal deep learning / Multimodal deep learning
- multiple layer model / Multiple layer model
N
- natural image datasets
- about / Natural image datasets
- batch normalization / Batch normalization
- global average pooling / Global average pooling
- natural language performance
- metrics for / Metrics for natural language performance
- Natural Language Processing (NLP) / Sequence-to-sequence networks for natural language processing
- negative particles / Restricted Boltzmann Machines
- Nesterov Accelerated Gradient / Optimization and other update rules
- network input normalization / Batch normalization
- Neural Machine Translation (NMT) / Weight tying
- Neural Network Language Models (NNLM) / Weight tying
- Neural Turing Machines (NTM)
- retrieve information in / Store and retrieve information in Neural Turing Machines
- store information in / Store and retrieve information in Neural Turing Machines
- about / Store and retrieve information in Neural Turing Machines
O
- off-policy training / Training stability
- Online training / Training stability
- Open-AI Gym
- about / Simulation environments
- URL / Simulation environments
- optimal state value function v(s) / Q-learning
- optimization / Optimization and other update rules
- out-of-vocabulary (OOV) / Preprocessing text data
P
- Part of Speech (POS) / Applications of RNN
- Platoon
- reference link / Multi-GPU
- policy gradients (PG)
- about / Policy gradients with REINFORCE algorithms
- with REINFORCE algorithms / Policy gradients with REINFORCE algorithms
- policy network / Policy gradients with REINFORCE algorithms
- positive and negative phases / Restricted Boltzmann Machines
- predictions
- example / Example of predictions
- Principal Component Analysis (PCA) / Visualizing the learned embeddings
Q
- Q-learning / Q-learning
- quantitative analysis / Evaluating embeddings – quantitative analysis
R
- recurrent highway networks (RHN) / Recurrent Highway Networks
- Recurrent Neural Network (RNN) / Encoding and embedding
- recurrent neural networks (RNN)
- need for / Need for RNN
- about / Need for RNN
- applications / Applications of RNN
- reduction operators / Reduction operators
- region-based localisation networks / Region-based localization networks
- Region Proposal Network (RPN) / Region-based localization networks
- reinforcement learning tasks / Reinforcement learning tasks
- replay memory / Training stability
- residual block / Residual connections
- residual connections / Residual connections
- residuals / Residual connections
- Restricted Boltzmann Machine / Restricted Boltzmann Machines
- RMSProp / Optimization and other update rules
- RNN
- dropout / Dropout for RNN
- deep approaches / Deep approaches for RNN
S
- SegNet network / Deconvolutions for images
- semi-supervised learning / Semi-supervised learning
- sequence-to-sequence (Seq2seq) network
- for natural language processing / Sequence-to-sequence networks for natural language processing
- about / Sequence-to-sequence networks for natural language processing, Seq2seq for translation
- for translation / Seq2seq for translation
- for chatbots / Seq2seq for chatbots
- efficiency, improving / Improving efficiency of sequence-to-sequence network
- SharpMask / Deconvolutions for images
- simple recurrent network
- about / Simple recurrent network
- LSTM network / LSTM network
- gated recurrent network / Gated recurrent network
- simulation environments / Simulation environments
- single-layer linear model / Single-layer linear model
- Single Instruction Multiple Data (SIMD) / Theano Op in C for GPU
- spatial transformer networks (STN) / A localization network
- stability
- training / Training stability
- stacked recurrent networks / Stacked recurrent networks
- state-action value network / Deep Q-network
- state value network / Policy gradients with REINFORCE algorithms
- state values / Q-learning
- stochastic depth / Stochastic depth
- stochastic gradient descent (SGD) / Backpropagation and stochastic gradient descent, Optimization and other update rules
- Streaming Multiprocessors (SM) / Theano Op in C for GPU
- symbolic computing / Graphs and symbolic computing
- loops in / Loops in symbolic computing
T
- t-distributed Stochastic Neighbor Embedding (t-SNE) / Visualizing the learned embeddings
- Tensor Processing Units (TPU) / Model conversions
- tensors
- need for / The need for tensors
- about / Tensors
- operations on / Operations on tensors
- dimension manipulation operators / Dimension manipulation operators
- elementwise operators / Elementwise operators
- reduction operators / Reduction operators
- linear algebra operators / Linear algebra operators
- text data
- preprocessing / Preprocessing text data
- Theano
- installing / Installing and loading Theano
- loading / Installing and loading Theano
- conda package / Conda package and environment manager
- environment manager / Conda package and environment manager
- installing, on CPU / Installing and running Theano on CPU
- executing, on CPU / Installing and running Theano on CPU
- GPU drivers / GPU drivers and libraries
- GPU libraries / GPU drivers and libraries
- installing, on GPU / Installing and running Theano on GPU
- executing, on GPU / Installing and running Theano on GPU
- debugging / Configuration, profiling and debugging
- profiling / Configuration, profiling and debugging
- configuration / Configuration, profiling and debugging
- Theano Op
- in Python, for CPU / Theano Op in Python for CPU
- in Python, for GPU / Theano Op in Python for the GPU
- in C, for CPU / Theano Op in C for CPU
- in C for GPU / Theano Op in C for GPU
- Torcs
- URL / Simulation environments
- training program
- structure / Structure of a training program
- script environment, setting up / Structure of a training program
- data, loading / Structure of a training program
- data, preprocessing / Structure of a training program
- model, building / Structure of a training program
- training / Structure of a training program
U
- unsupervised learning
- with co-localisation / Unsupervised learning with co-localization
- update rules / Optimization and other update rules
V
- validation dataset
- training / Training
- variables / Memory and variables
- variational RNN / Dropout for RNN
W
- weight tying (WT) / Weight tying
- word embeddings
- application / Application of word embeddings
- Word Error Rate (WER) / Metrics for natural language performance
Y
- You Only See Once (YOLO) architecture / Region-based localization networks