Machine Learning with Artificial Neural Networks in Clojure
summary
Presents a hand-implemented artificial neural network library written in Clojure using Clojure protocols to encode forward and backpropagation equations, plus a tiny DSL for defining networks. Demonstrates training an XOR network to under 1% error in a few iterations and describes an MNIST workflow (60,000 training / 10,000 test images) mentioning convolutional nets achieving about 0.23% error. Also describes packaging as a single JAR server with an unfinished ZeroMQ pub/sub idea, a Go client that POSTs images, and audible training-progress notifications.
AI-written summary · about
chapters (8)
- 0:00AI and neural networks
- 5:02Why neural networks work
- 13:47Neurons and perceptrons
- 27:57Neural networks in Clojure
- 31:55XOR network demo
- 35:11MNIST server demo
- 39:25ANN architecture and deployment
- 43:06Summary
related talks
29mDeep Learning Needs Clojure
28mMachine Learning Live
40mBuilding Machine Learning Models with Clojure and Cortex
44mMachine Learning Exposed
51mMachine Learning Exposed Deep and Reinforcement Learning
38mDeclarative Deep Learning In Clojure
26mMachine Learning in the Browser with Deep Neural Networks
26mThe Feynman Machine: AI architecture in Clojure/Neanderthal