On the State of the Art of Evaluation in Neural Language Models

Abstract

Ongoing innovations in recurrent neural network architectures have provided a steady influx of apparently state-of-the-art results on language modelling benchmarks. However, these have been evaluated using differing code bases and limited computational resources, which represent uncontrolled sources of experimental variation. We reevaluate several popular architectures and regularisation methods with large-scale automatic black-box hyperparameter tuning and arrive at the somewhat surprising conclusion that standard LSTM architectures, when properly regularised, outperform more recent models. We establish a new state of the art on the Penn Treebank and Wikitext-2 corpora, as well as strong baselines on the Hutter Prize dataset.

References in corpus (3)

• Neural Architecture Search with Reinforcement Learning
• Pointer Sentinel Mixture Models
• Dynamic Evaluation of Neural Sequence Models

Cited by in corpus (7)

• Deep learning for molecular design - a review of the state of the art
• Gmail Smart Compose: Real-Time Assisted Writing
• Neural Text Generation: A Practical Guide
• Differentiable Architecture Search with Ensemble Gumbel-Softmax
• WeNet: Weighted Networks for Recurrent Network Architecture Search
• Variational Smoothing in Recurrent Neural Network Language Models
• Measuring the Complexity of Domains Used to Evaluate AI Systems