Coded-InvNet for Resilient Prediction Serving Systems
arXiv:2106.06445
Abstract
Inspired by a new coded computation algorithm for invertible functions, we propose Coded-InvNet a new approach to design resilient prediction serving systems that can gracefully handle stragglers or node failures. Coded-InvNet leverages recent findings in the deep learning literature such as invertible neural networks, Manifold Mixup, and domain translation algorithms, identifying interesting research directions that span across machine learning and systems. Our experimental results show that Coded-InvNet can outperform existing approaches, especially when the compute resource overhead is as low as 10%. For instance, without knowing which of the ten workers is going to fail, our algorithm can design a backup task so that it can correctly recover the missing prediction result with an accuracy of 85.9%, significantly outperforming the previous SOTA by 32.5%.
References in corpus (10)
- Distilling the Knowledge in a Neural Network
- Conditional Generative Adversarial Nets
- Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
- mixup: Beyond Empirical Risk Minimization
- An Overview of Multi-Task Learning in Deep Neural Networks
- Speeding Up Distributed Machine Learning Using Codes
- Manifold Mixup: Better Representations by Interpolating Hidden States
- Adversarial Robustness as a Prior for Learned Representations
- Learning a Code: Machine Learning for Approximate Non-Linear Coded Computation
- The Lipschitz Constant of Self-Attention