Machine Learning for Gophers
summary
Claim is that Go can do ML today but is not yet ideal: the talk reviews gonum, gorgonia, tensorflow and implements logistic regression on MNIST using gonum and a custom mat package. Iteration time drops from about 15s to 7s to 1s after goroutine parallelism and assembly/BLAS optimizations; a tiny subset yields 100% on tiny subset while the full run reaches only 39% accuracy after 30s. Notes cover mutability pitfalls, 60k goroutines, iTerm image plotting, and that tensorflow Go currently does inference not training.
AI-written summary · about
chapters (8)
- 0:00Start
- 10:36What is Machine Learning?
- 11:08Input
- 12:30Logistic regression
- 14:56What do we need?
- 18:01Let's reinvent the wheel
- 27:18References
- 28:24The future
related talks
18mMachine Learning and Go
50mDeep Learning with Go
33mspaGO: a work-in-progress yet ready-to-use library for ML/NLP
37mA Funny Thing Happened On The Way To Reimplementing AlphaGo
29mThe Gopher Orchestra: Distributed Machine Learning with Go
29mThe Go-to Language for AI: Opportunities and Challenges
26mDeep learning in Go - Present and future
04mGo and data, can they ever be friends?