Learning to Learn by Zeroth-Order Oracle
This work addresses optimization challenges in scenarios with limited gradient access, such as black-box adversarial attacks, representing an incremental advancement in learning-to-learn methods.
The paper tackles the problem of zeroth-order optimization where gradient information is unavailable, by extending the learning-to-learn framework to learn an optimizer that approximates gradients and reduces variance, resulting in improved convergence rates and final solutions on tasks like black-box adversarial attacks.
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.