Machine Learning Live
summary
A Clojure-based startup toolkit is presented that runs on the JVM, supports REPL-driven interactive development, and uses backprop-trained neural networks with live visualizations. Concrete demos include a Scrabble mapping solved to 26/26 by encoding letters as 26-bit inputs and scores as 4-bit outputs, and a handwritten digit pipeline that uses an autoencoder compressor 784→150 trained on 60,000 examples and tested on 10,000, achieving about 90% accuracy (≈10% error) with interpretable feature visualizations of learned weights.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 0:59What is machine learning
- 2:18Building functions from experience
- 3:17Machine Learning vs Flight
- 3:59New
- 4:39Why Java
- 7:12Neural networks
- 8:26Training neural networks
- 9:35Scrabble
- 10:02Code
- 15:56Handwritten Digit Recognition
- 16:45Neural Network Compression
- 18:22Function Composition
- 19:01Compression
- 27:04Outro
related talks
29mDeep Learning Needs Clojure
44mMachine Learning with Artificial Neural Networks in Clojure
42mLearning Machine Learning
25mApplied Machine Learning in Elixir
51mMachine Learning Exposed Deep and Reinforcement Learning
44mMachine Learning Exposed
38mMachine Learning with Clojure and Apache Spark- 42mLearning Machine Learning