A Study on Encodings for Neural Architecture Search
Neural architecture search (NAS) has been extensively studied in the past few years. A popular approach is to represent each neural architecture in the search space as a directed acyclic graph (DAG), and then search over all DAGs by encoding the adjacency matrix and list of operations as a set of hyperparameters. Recent work has demonstrated that even small changes to the way each architecture is encoded can have a significant effect on the performance of NAS algorithms. In this work, we present the first formal study on the effect of architecture encodings for NAS, including a theoretical grounding and an empirical study. First we formally define architecture encodings and give a theoretical characterization on the scalability of the encodings we study Then we identify the main encoding-dependent subroutines which NAS algorithms employ, running experiments to show which encodings work best with each subroutine for many popular algorithms. The experiments act as an ablation study for prior work, disentangling the algorithmic and encoding-based contributions, as well as a guideline for future work. Our results demonstrate that NAS encodings are an important design decision which can have a significant impact on overall performance. Our code is available at https://github.com/naszilla/nas-encodings.
Code (2)
Tasks
Neural Architecture SearchSimilar Papers 제목 키워드 기반
Encodings for Prediction-based Neural Architecture Search
Predictor-based methods have substantially enhanced Neural Architecture Search (NAS) optimization. The efficacy of these predictors is largely influenced by the method of encoding neural network architectures. While trad…
Neural Architecture SearchPredictionTransfer LearningCATE: Computation-aware Neural Architecture Encoding with Transformers
Recent works (White et al., 2020a; Yan et al., 2020) demonstrate the importance of architecture encodings in Neural Architecture Search (NAS). These encodings encode either structure or computation information of the neu…
AutoMLNeural Architecture SearchRepresentation LearningUnsupervised Pre-trainingTowards Code-Oriented LM Embeddings for Surrogate-Assisted Neural Architecture Search
Developing effective surrogates (performance predictors) for Neural Architecture Search (NAS) typically requires expensive fine-tuning or the engineering of complex representations. We propose a low-cost embedding strate…
Neural Architecture SearchFunctional Equivalence in Attention: A Comprehensive Study with Applications to Linear Mode Connectivity
Neural network parameter spaces are inherently non-injective, as distinct parameter configurations can realize identical functions through functional equivalence. While this symmetry is well understood in classical fully…
Encoding Selection for Solving Hamiltonian Cycle Problems with ASP
It is common for search and optimization problems to have alternative equivalent encodings in ASP. Typically none of them is uniformly better than others when evaluated on broad classes of problem instances. We claim tha…