Trellis Networks for Sequence Modeling

Abstract

We present trellis networks, a new architecture for sequence modeling. On the one hand, a trellis network is a temporal convolutional network with special structure, characterized by weight tying across depth and direct injection of the input into deep layers. On the other hand, we show that truncated recurrent networks are equivalent to trellis networks with special sparsity structure in their weight matrices. Thus trellis networks with general weight matrices generalize truncated recurrent networks. We leverage these connections to design high-performing trellis networks that absorb structural and algorithmic elements from both recurrent and convolutional models. Experiments demonstrate that trellis networks outperform the current state of the art methods on a variety of challenging benchmarks, including word-level language modeling and character-level language modeling tasks, and stress tests designed to evaluate long-term memory retention. The code is available at https://github.com/locuslab/trellisnet .

Published at ICLR 2019

References in corpus (11)

• Sequence to Sequence Learning with Neural Networks
• Natural Language Processing (almost) from Scratch
• An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling
• Neural Architecture Search with Reinforcement Learning
• WaveNet: A Generative Model for Raw Audio
• DARTS: Differentiable Architecture Search
• On the Properties of Neural Machine Translation: Encoder-Decoder Approaches
• Neural Machine Translation in Linear Time
• An Analysis of Neural Language Modeling at Multiple Scales
• Improving Neural Language Models with a Continuous Cache
• Stable Recurrent Models

Cited by in corpus (23)

• Efficiently Modeling Long Sequences with Structured State Spaces
• A Survey of Sound Source Localization with Deep Learning Methods
• SCINet: Time Series Modeling and Forecasting with Sample Convolution and Interaction
• HiPPO: Recurrent Memory with Optimal Polynomial Projections
• RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
• Multiscale Deep Equilibrium Models
• Compressive Transformers for Long-Range Sequence Modelling
• Compression of Recurrent Neural Networks for Efficient Language Modeling
• Temporal Convolutional Attention-based Network For Sequence Modeling
• Parameter Efficient Multimodal Transformers for Video Representation Learning
• Addressing Some Limitations of Transformers with Feedback Memory
• Improving the Gating Mechanism of Recurrent Neural Networks
• FlexConv: Continuous Kernel Convolutions with Differentiable Kernel Sizes
• CKConv: Continuous Kernel Convolution For Sequential Data
• NAC-TCN: Temporal Convolutional Networks with Causal Dilated Neighborhood Attention for Emotion Understanding
• A global convergence theory for deep ReLU implicit networks via over-parameterization
• Time Distributed Deep Learning Models for Purely Exogenous Forecasting: Application to Water Table Depth Prediction using Weather Image Time Series
• Behavior Gated Language Models
• Rethinking Neural Operations for Diverse Tasks
• The Uncanny Similarity of Recurrence and Depth
• Lightweight, Dynamic Graph Convolutional Networks for AMR-to-Text Generation
• Constructing Gradient Controllable Recurrent Neural Networks Using Hamiltonian Dynamics
• CommUnet: U-net decoder for convolutional codes in communication