Summary of Unlocking Fednl: Self-contained Compute-optimized Implementation, by Konstantin Burlachenko et al.
Unlocking FedNL: Self-Contained Compute-Optimized Implementation
by Konstantin Burlachenko, Peter Richtárik
First submitted to arxiv on: 11 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Mathematical Software (cs.MS); Performance (cs.PF); Optimization and Control (math.OC)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper introduces Federated Newton Learn (FedNL) algorithms, a significant step towards applying second-order methods to Federated Learning (FL) and large-scale optimization. The authors address three practical drawbacks of the reference FedNL prototype: slow launch times, limited simulation capabilities, and challenging integration into resource-constrained applications. To bridge this gap, they present a self-contained implementation of FedNL for single-node and multi-node settings. This implementation reduces wall clock time by x1000 and outperforms alternatives in training logistic regression using CVXPY, Apache Spark, Ray/Scikit-Learn. The paper also proposes two practical-oriented compressors for FedNL: adaptive TopLEK and cache-aware RandSeqK. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper improves Federated Learning (FL) by introducing a family of algorithms called Federated Newton Learn (FedNL). FL allows machines to learn together without sharing their data. The current version of FedNL has some issues, like taking too long to start and not being able to work with many devices at once. To fix these problems, the authors created a new version that works faster and can be used in different situations. |
Keywords
» Artificial intelligence » Federated learning » Logistic regression » Optimization
