Apache SystemML: State of the Project and Future Plans
summary
Project is now a top-level Apache offering and ships R and Python APIs plus an experimental Scala DSL that compiles matrix-based code into execution plans. A lossless compressed linear algebra engine reduces memory footprint and yields 5–26× faster end-to-end training in benchmarks (including compression overhead). Dynamic code generation creates fused kernels that beat many hand-tuned implementations. Deep learning gains an importer for Caffe/Keras models and pluggable GPU backends, delivering roughly an order‑of‑magnitude speedup on some inference workloads.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:28What is Apache SystemML
- 7:12What is new in SystemML
- 11:30Compressed Linear Algebra
- 17:05fused kernels
- 21:35deep learning
- 25:00conclusion
related talks
16mAndrew Musselman, Committer & PMC Member, Apache Mahout at MLconf Seattle 2017
39mMachine Learning with Scala on Spark
42mWhy Choose One: Multi-Engine Analytics with Apache Wayang
44mJava for AI
36mMachine Learning with Apache Beam
16mAmeet Talwalkar, Assistant Professor of Computer Science, UCLA @ MLconf SF Part 2
42mBridge the chasm between your ML and app devs with Semantic Kernel
02mAkashML