Private Adaptive Optimization with Side Information

by   Tian Li, et al.

Adaptive optimization methods have become the default solvers for many machine learning tasks. Unfortunately, the benefits of adaptivity may degrade when training with differential privacy, as the noise added to ensure privacy reduces the effectiveness of the adaptive preconditioner. To this end, we propose AdaDPS, a general framework that uses non-sensitive side information to precondition the gradients, allowing the effective use of adaptive methods in private settings. We formally show AdaDPS reduces the amount of noise needed to achieve similar privacy guarantees, thereby improving optimization performance. Empirically, we leverage simple and readily available side information to explore the performance of AdaDPS in practice, comparing to strong baselines in both centralized and federated settings. Our results show that AdaDPS improves accuracy by 7.7 privacy-utility trade-offs on large-scale text and image benchmarks.


Differentially Private Adaptive Optimization with Delayed Preconditioners

Privacy noise may negate the benefits of using adaptive optimizers in di...

Private Multi-Task Learning: Formulation and Applications to Federated Learning

Many problems in machine learning rely on multi-task learning (MTL), in ...

Accelerated Federated Learning with Decoupled Adaptive Optimization

The federated learning (FL) framework enables edge clients to collaborat...

Mixed Differential Privacy in Computer Vision

We introduce AdaMix, an adaptive differentially private algorithm for tr...

Large Scale Private Learning via Low-rank Reparametrization

We propose a reparametrization scheme to address the challenges of apply...

Learning under Storage and Privacy Constraints

Storage-efficient privacy-guaranteed learning is crucial due to enormous...

On Dynamic Noise Influence in Differentially Private Learning

Protecting privacy in learning while maintaining the model performance h...

Please sign up or login with your details

Forgot password? Click here to reset