More Clojure Less Complication
summary
Describes Clojure-native ML stacks: Deep Diamond for tensors and Neanderthal for matrices, with CUDA/OpenCL GPU backends so the same Clojure code runs on CPU or GPU. Shows a regression example with a small network of ~5,000 parameters that trains in under a second with accelerators (versus hours without). Highlights branching/gradient support now in snapshots, the libraries total under 9,000 lines of code, are EPL-licensed, and are accompanied by books that teach building and extending the tools.
AI-written summary · about
chapters (6)
- 0:00Start
- 8:06LOAD THE DATA
- 12:53CREATE THE NEURAL NETWORK
- 19:00CONVOLUTIONAL NETWORKS
- 20:33DIRECTED ACYCLICAL GRAPHS
- 24:09SUMMARY
related talks
41mInteractive GPU programming with ClojureCUDA & ClojureCL
42mClojure is Not Afraid of the GPU
29mDeep Learning Needs Clojure
40mThe Tensors Must Flow
26mThe Feynman Machine: AI architecture in Clojure/Neanderthal
40mBuilding Machine Learning Models with Clojure and Cortex
38mDeclarative Deep Learning In Clojure
41mDeploying ML models in a Clojure environment