Declarative Deep Learning In Clojure
summary
A declarative wrapper brings Deeplearning4j into Clojure by letting you pass keyword args that map to Java builders and returning pure Clojure data structures for easy testing and refactoring. It supports feedforward, CNN and LSTM layers, early stopping, Spark training and evaluation. Example MNIST runs report MNIST 0.98 accuracy and an LSTM example around 0.95 accuracy, and models train on laptop in ~40 minutes without GPU. Planned additions include CUDA, reinforcement learning and specs.
AI-written summary · about
chapters (28)
- 0:00Intro
- 0:22About me
- 0:53Yet Analytics
- 1:19The Matrix
- 1:45Optimization
- 2:40Neurons
- 5:04Networks
- 5:41Associative Learning
- 6:52Demonstration
- 7:42Non Associative
- 8:19Habituation
- 9:30sensitization
- 10:00combined properties
- 11:49model of memory
- 13:09feedforward neural networks
- 13:51adaptive resonance theory
- 16:33LSTM
- 16:43Understanding LS TMS
- 19:30Java Builder Pattern
- 19:49Constructors
- 21:56Builder Pattern
- 24:34Java Interop
- 26:32Live Demo
- 29:14Data Structure
- 30:22Accuracy
- 31:04LSDM Example
- 33:26Whats Next
- 34:11Jason Lewis
related talks
50mApplying Machine Learning to IoT Use Cases with Deeplearning4j
25mMore Clojure Less Complication
28mMachine Learning Live
38mMachine Learning with Clojure and Apache Spark
44mMachine Learning with Artificial Neural Networks in Clojure
44mDeep Learning in Production
35mDeconstructing Deep Learning
50mBuilding Deep Neural Networks using Deep Learning for Java(DL4J)