Neural architecture search
Collections
https://www.ml4aad.org/automl/literature-on-neural-architecture-search
Neural Architecture Search: A Survey
https://arxiv.org/pdf/1808.05377.pdf
Search space
Small search space reduces search time, but introduces a human bias, and prevents finding novel architecture
Search strategy
How to explore the search space
Performance estimation strategy
The objectives of NAS is typically to find architectures that achieve high predictive performance on unseen data.
Search space
Input of layer i can be formally described as a function:
Chain-structured network:
Residual type network:
Dense connected network:
Search strategy
Random search
Bayesian optimization
Evolutionary methods
Reinforcement learning
Gradient based methods
Performance estimation strategy
Regular training
Lower resolution image
Less filter per layer
Network morphism
One-shot architecture search
Neural Architecture Search with Reinforcement Learning
https://arxiv.org/pdf/1611.01578.pdf
AutoAugment: Learning Augmentation Policies from Data
Last updated