Learning to Learn by Zeroth-Order Oracle
arXiv:1910.09464
Abstract
In the learning to learn (L2L) framework, we cast the design of optimization algorithms as a machine learning problem and use deep neural networks to learn the update rules. In this paper, we extend the L2L framework to zeroth-order (ZO) optimization setting, where no explicit gradient information is available. Our learned optimizer, modeled as recurrent neural network (RNN), first approximates gradient by ZO gradient estimator and then produces parameter update utilizing the knowledge of previous iterations. To reduce high variance effect due to ZO gradient estimator, we further introduce another RNN to learn the Gaussian sampling rule and dynamically guide the query direction sampling. Our learned optimizer outperforms hand-designed algorithms in terms of convergence rate and final solution on both synthetic and practical ZO optimization tasks (in particular, the black-box adversarial attack task, which is one of the most widely used tasks of ZO optimization). We finally conduct extensive analytical experiments to demonstrate the effectiveness of our proposed optimizer.
Published as a conference paper at ICLR 2020
References in corpus (10)
- ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Models
- Learning to Learn without Gradient Descent by Gradient Descent
- Learning to Optimize
- Learned Optimizers that Scale and Generalize
- Query-Efficient Hard-label Black-box Attack:An Optimization-based Approach
- Zeroth-Order Stochastic Variance Reduction for Nonconvex Optimization
- Learning to Optimize Neural Nets
- A Comprehensive Linear Speedup Analysis for Asynchronous Stochastic Parallel Optimization from Zeroth-Order to First-Order
- Stochastic Zeroth-order Optimization via Variance Reduction method
- Zeroth-order Asynchronous Doubly Stochastic Algorithm with Variance Reduction