ML in Java, YES it's possible!
summary
Three Java paths are shown and compared: VISR / JSR 381 (compact image-classification API with DPNet default, trains a small convolutional net from JSON and exports .dnet), DJL (Deep Java Library) (multi-backend model server via Docker, model zoo and serving, supports PyTorch/TensorFlow/MXNet), and Deeplearning4J (DL4J) plus Tribuo for classical ML. Training runs locally on an M1 CPU with about 100 images; demo model classifies hot-dog vs not-hot-dog at a 50% threshold and the inference works end-to-end.
AI-written summary · about
chapters (13)
- 0:00Introduction to ML in Java
- 1:26Why Java for machine learning
- 5:23Traditional vs ML programming
- 8:55Machine learning basics
- 12:51Deep learning explained
- 15:48Challenges with ML in Java
- 17:38Java ML frameworks overview
- 20:04Demo: VISR image classification
- 28:46Deep Learning 4J framework
- 38:03AWS Deep Java Library (DGL)
- 43:01Oracle Tribo for traditional ML
- 45:30End-to-end ML in Java
- 46:14Q&A session
related talks
46mML in Java, YES it's possible!
47mML in Java, YES it's possible!
01h 01mDeep Learning and Java
58mPatterns, Predictions and Programming
25mDeeplearning4j: How to run AI workloads on Jakarta EE compliant servers
43mWhy Java and the JVM Will Dominate the Future of Machine Learning, AI, and Big Data
48mML on the JVM
29mAchieve real-time inference at scale on AWS with Deep Java Library