Skip to content

Latest commit

 

History

History
24 lines (21 loc) · 2.87 KB

Pasunuru2020FENAS.md

File metadata and controls

24 lines (21 loc) · 2.87 KB

Title

FENAS: Flexible and Expressive Neural Architecture Search

Author

Ramakanth Pasunuru, Mohit Bansal

Abstract

Architecture search is the automatic process of designing the model or cell structure that is optimal for the given dataset or task. Recently, this approach has shown good improvements in terms of performance (tested on language modeling and image classification) with reasonable training speed using a weight sharing-based approach called Efficient Neural Architecture Search (ENAS). In this work, we propose a novel architecture search algorithm called Flexible and Expressible Neural Architecture Search (FENAS), with more flexible and expressible search space than ENAS, in terms of more activation functions, input edges, and atomic operations. Also, our FENAS approach is able to reproduce the well-known LSTM and GRU architectures (unlike ENAS), and is also able to initialize with them for finding architectures more efficiently. We explore this extended search space via evolutionary search and show that FENAS performs significantly better on several popular text classification tasks and performs similar to ENAS on standard language model benchmark. Further, we present ablations and analyses on our FENAS approach.

Bib

@inproceedings{pasunuru-bansal-2020-fenas, title = "{FENAS}: Flexible and Expressive Neural Architecture Search", author = "Pasunuru, Ramakanth and Bansal, Mohit", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2020", month = nov, year = "2020", address = "Online", publisher = "Association for Computational Linguistics", url = "https://www.aclweb.org/anthology/2020.findings-emnlp.258", doi = "10.18653/v1/2020.findings-emnlp.258", pages = "2869--2876", abstract = "Architecture search is the automatic process of designing the model or cell structure that is optimal for the given dataset or task. Recently, this approach has shown good improvements in terms of performance (tested on language modeling and image classification) with reasonable training speed using a weight sharing-based approach called Efficient Neural Architecture Search (ENAS). In this work, we propose a novel architecture search algorithm called Flexible and Expressible Neural Architecture Search (FENAS), with more flexible and expressible search space than ENAS, in terms of more activation functions, input edges, and atomic operations. Also, our FENAS approach is able to reproduce the well-known LSTM and GRU architectures (unlike ENAS), and is also able to initialize with them for finding architectures more efficiently. We explore this extended search space via evolutionary search and show that FENAS performs significantly better on several popular text classification tasks and performs similar to ENAS on standard language model benchmark. Further, we present ablations and analyses on our FENAS approach.", }