Blocks and Fuel: Frameworks for deep learning

Abstract

We introduce two Python frameworks to train neural networks on large datasets: Blocks and Fuel. Blocks is based on Theano, a linear algebra compiler with CUDA-support. It facilitates the training of complex neural network models by providing parametrized Theano operations, attaching metadata to Theano's symbolic computational graph, and providing an extensive set of utilities to assist training the networks, e.g. training algorithms, logging, monitoring, visualization, and serialization. Fuel provides a standard format for machine learning datasets. It allows the user to easily iterate over large datasets, performing many types of pre-processing on the fly.

References in corpus (4)

• On the difficulty of training Recurrent Neural Networks
• Theano: new features and speed improvements
• DRAW: A Recurrent Neural Network For Image Generation
• Pylearn2: a machine learning research library

Cited by in corpus (39)

• Attention-Based Models for Speech Recognition
• Adversarially Learned Inference
• Semi-Supervised Learning with Ladder Networks
• An Actor-Critic Algorithm for Sequence Prediction
• A Neural Autoregressive Approach to Collaborative Filtering
• Artificial Neural Networks Applied to Taxi Destination Prediction
• Boundary-Seeking Generative Adversarial Networks
• Dropout as data augmentation
• Improved Deep Learning Baselines for Ubuntu Corpus Dialogs
• A Tutorial on Deep Learning for Music Information Retrieval
• Syntactically Guided Neural Machine Translation
• Towards End-to-End Speech Recognition with Deep Convolutional Neural Networks
• Learning Language-Visual Embedding for Movie Understanding with Natural-Language
• Learning to Compute Word Embeddings On the Fly
• Invariant Representations for Noisy Speech Recognition
• Kapre: On-GPU Audio Preprocessing Layers for a Quick Implementation of Deep Neural Network Models with Keras
• Embracing data abundance: BookTest Dataset for Reading Comprehension
• A Survey on Deep Learning Methods for Robot Vision
• Regularizing RNNs by Stabilizing Activations
• Implementing graph neural networks with TensorFlow-Keras
• Task Loss Estimation for Sequence Prediction
• Deep Learning in Bioinformatics
• Convolutional RNN: an Enhanced Model for Extracting Features from Sequential Data
• TensorFlow Estimators: Managing Simplicity vs. Flexibility in High-Level Machine Learning Frameworks
• Bidirectional Helmholtz Machines
• Hybrid Dialog State Tracker
• nuts-flow/ml: data pre-processing for deep learning
• Dynamic Capacity Networks
• Why not be Versatile? Applications of the SGNMT Decoder for Machine Translation
• Deeply Learned Preselection of Higgs Dijet Decays at Future Lepton Colliders
• Action Classification and Highlighting in Videos
• An Efficient Character-Level Neural Machine Translation
• Read, Tag, and Parse All at Once, or Fully-neural Dependency Parsing
• Stochastic Sequential Neural Networks with Structured Inference
• Can deep learning help you find the perfect match?
• Semantics, Representations and Grammars for Deep Learning
• SeNA-CNN: Overcoming Catastrophic Forgetting in Convolutional Neural Networks by Selective Network Augmentation
• Plug and Play, Model-Based Reinforcement Learning
• On Multilingual Training of Neural Dependency Parsers