AlphaX: eXploring Neural Architectures with Deep Neural Networks and Monte Carlo Tree Search

By Linnan Wang, Yiyang Zhao, Yuu Jinnai. 2019

100 citations recorded by the source corpus.

We present AlphaX, a fully automated agent that designs complex neural architectures from scratch. AlphaX explores the exponentially grown search space with a distributed Monte Carlo Tree Search (MCTS) and a Meta-Deep Neural Network (DNN). MCTS intrinsically improves the search efficiency by dynamically balancing the exploration and exploitation at fine-grained states, while Meta-DNN predicts the network accuracy to guide the search, and to provide an estimated reward to speed up the rollout. As the search progresses, AlphaX also generates the training data for Meta-DNN. So, the learning of Meta-DNN is end-to-end. In 14 days with only 16 GPUs (1832 samples), AlphaX found an architecture that reaches the state-of-the-art accuracies on both CIFAR-10(97.18%) and ImageNet(75.5% top-1 and 92.2% top-5). This demonstrates up to 10x speedup over the original searching for NASNet that used 500 GPUs in 4 days (20000 samples). On NASBench-101, AlphaX demonstrates 3x and 2.8x speedup over Random Search and Regularized Evolution. Finally, we show the searched architecture improves a variety of vision applications from Neural Style Transfer, to Image Captioning and Object Detection. Our implementation is available at this https URL.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Neural Architecture Search & AutoML

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Neural Architecture Search & AutoML

Citations in the corpus

BANANAS: Bayesian Optimization with Neural Architectures for Neural Architecture Search (2019, methodology): “2016), REINFORCE (Williams 1992), GP-based BO (Snoek, Larochelle, and Adams 2012), AlphaX (Wang et al. 2018), ASHA (Li and Talwalkar 2019), GCN Predictor (Wen et al.”
Evaluating the Search Phase of Neural Architecture Search (2019, background): “Ying et al. (2019) introduced a dataset that contains the ground-truth performance of CNN cells, and Wang et al. (2019) evaluated some traditional search algorithms on it.”
Bridging the Gap between Sample-based and One-shot Neural Architecture Search with BONAS (2020, methodology): “Following [33, 34], Figure 4 shows the performance of the best model after using a given number of architecture samples.”
Neural Architecture Search with Reinforce and Masked Attention Autoregressive Density Estimators (2020, methodology): “We include random search, which is regarded as a competitive baseline (Li & Talwalkar, 2019), regularized evolution (Real et al., 2019), and AlphaX, an RL algorithm that uses MCTS (Wang et al., 2018).”
ELF OpenGo: An Analysis and Open Reimplementation of AlphaZero (2019, background): “The release of AGZ and AZ has motivated a line of preliminary work (Addanki et al., 2019; Dong et al., 2017; Wang & Rompf, 2018; Wang et al., 2018) which aims to analyze and understand the algorithms, as well as to apply similar algorithms to other domains.”
Local Search is State of the Art for NAS Benchmarks (2020, methodology): “On NASBench-101, we test local search with the aforementioned algorithms, as well as REINFORCE (Williams, 1992) and AlphaX (Wang et al., 2018).”
vPipe: A Virtualized Acceleration System for Achieving Efficient and Scalable Pipeline Parallel DNN Training (2022, background): “Recently, more and more developers have adopted dynamic DNN training where the number of layers varies with the training inputs (e.g., DyNet [34]) or the training is exploratory (e.g., neural architecture search [45], [55], [57], [62]).”
Multi-objective Neural Architecture Search via Predictive Network Performance Optimization (2019, unknown): “Compared with MLP and LSTM predictors proposed before (Wang et al., 2019b), GCN can preserve the context of graph data better.”
Optimal Transport Kernels for Sequential and Parallel Neural Architecture Search (2020, methodology): “We do not compare with Reinforcement Learning approaches (Pham et al., 2018) and AlphaX (Wang et al., 2020) which have been shown to perform poorly in (White et al., 2021).”
Local Search is State of the Art for Neural Architecture Search Benchmarks (2020, methodology): “On NASBench-101, we test local search with the aforementioned algorithms, as well as REINFORCE [37] and AlphaX [33].”
Improving One-Shot NAS by Suppressing the Posterior Fading (2019, unknown): “Early neural architecture search (NAS) [24, 22, 29, 39, 2, 35] methods normally involves reinforcement learning or neuro-evolution.”