MONAS: Multi-Objective Neural Architecture Search using Reinforcement Learning
arXiv:1806.10332
Abstract
Recent studies on neural architecture search have shown that automatically designed neural networks perform as good as expert-crafted architectures. While most existing works aim at finding architectures that optimize the prediction accuracy, these architectures may have complexity and is therefore not suitable being deployed on certain computing environment (e.g., with limited power budgets). We propose MONAS, a framework for Multi-Objective Neural Architectural Search that employs reward functions considering both prediction accuracy and other important objectives (e.g., power consumption) when searching for neural network architectures. Experimental results showed that, compared to the state-ofthe-arts, models found by MONAS achieve comparable or better classification accuracy on computer vision applications, while satisfying the additional objectives such as peak power.
References in corpus (3)
Cited by in corpus (42)
- AutoML: A Survey of the State-of-the-Art
- Efficient Deep Learning: A Survey on Making Deep Learning Models Smaller, Faster, and Better
- Object Detection in 20 Years: A Survey
- Deep Reinforcement Learning for Multi-objective Optimization
- ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware
- A Survey on Neural Architecture Search
- Graph HyperNetworks for Neural Architecture Search
- Multi-Objective Hyperparameter Optimization in Machine Learning -- An Overview
- Fast, Accurate and Lightweight Super-Resolution with Neural Architecture Search
- NSGA-Net: Neural Architecture Search using Multi-Objective Genetic Algorithm
- A Comprehensive Survey on Hardware-Aware Neural Architecture Search
- Multi-Objective Reinforced Evolution in Mobile Neural Architecture Search
- Layer-compensated Pruning for Resource-constrained Convolutional Neural Networks
- CryptoNAS: Private Inference on a ReLU Budget
- Learning the Pareto Front with Hypernetworks
- Weight-Sharing Neural Architecture Search: A Battle to Shrink the Optimization Gap
- CARS: Continuous Evolution for Efficient Neural Architecture Search
- Machine Learning for the Detection and Identification of Internet of Things (IoT) Devices: A Survey
- Multi-Objective Evolutionary Design of Deep Convolutional Neural Networks for Image Classification
- Evolutionary Neural Architecture Search Supporting Approximate Multipliers
- Orchestrating the Development Lifecycle of Machine Learning-Based IoT Applications: A Taxonomy and Survey
- Learnable Embedding Space for Efficient Neural Architecture Compression
- Differentiable Architecture Search with Ensemble Gumbel-Softmax
- Can weight sharing outperform random architecture search? An investigation with TuNAS
- Learning to infer: RL-based search for DNN primitive selection on Heterogeneous Embedded Systems
- AttendNets: Tiny Deep Image Recognition Neural Networks for the Edge via Visual Attention Condensers
- InstaNAS: Instance-aware Neural Architecture Search
- Multi-objective Neural Architecture Search via Non-stationary Policy Gradient
- How can AI Automate End-to-End Data Science?
- Hardware Aware Neural Network Architectures using FbNet
- Single-Path Mobile AutoML: Efficient ConvNet Design and NAS Hyperparameter Optimization
- EdgeSegNet: A Compact Network for Semantic Segmentation
- Searching for Fast Model Families on Datacenter Accelerators
- AttendSeg: A Tiny Attention Condenser Neural Network for Semantic Segmentation on the Edge
- MemNet: Memory-Efficiency Guided Neural Architecture Search with Augment-Trim learning
- Joint Channel and Weight Pruning for Model Acceleration on Moblie Devices
- FLASH: Fast Neural Architecture Search with Hardware Optimization
- TND-NAS: Towards Non-differentiable Objectives in Progressive Differentiable NAS Framework
- Network Space Search for Pareto-Efficient Spaces
- Optimising the Performance of Convolutional Neural Networks across Computing Systems using Transfer Learning
- Conditional Neural Architecture Search
- The Efficiency Misnomer