The Gopher Orchestra: Distributed Machine Learning with Go
summary
Shows a working Go pipeline that shards data and trains models across nodes using goroutines and channels and demonstrates both Iris (150 rows) and wine datasets plus a Kaggle housing example. Implements pipeline, master‑worker, ring all‑reduce and parameter‑server patterns, uses gRPC for node comms and checkpointing for fault tolerance. Reports concrete results: KNN 94.7% versus decision tree 82–84%, logs epochs with loss decreasing and a runnable demo that prints stage logs and final accuracies.
AI-written summary · about
chapters (10)
- 0:00Introduction to talk
- 1:24Machine learning basics
- 4:54Types of machine learning
- 8:05Distributed machine learning explained
- 9:48Why Go for distributed ML
- 12:15Design patterns for ML systems
- 14:20Demo pipeline design pattern
- 19:32Building blocks of distributed ML
- 23:26Challenges in Go ML systems
- 27:15Final thoughts and resources
