The state of Scala as a platform for statistical computing
summary
Claim: Scala can be a first‑class environment for statistical computing but only with concrete pieces in place. Reimplementing a Python data munging job shows 3–4× memory overhead in naive Scala code; ongoing work produces a saddle frame for heterogeneously typed columnar data, spire numeric abstractions for generic numbers, and a Scala notebook plus D3 for visualization. Heterogeneous record typing still needs shapeless glue, and notebook UX and library coordination remain unfinished.
AI-written summary · about
chapters (16)
- 0:00Start
- 3:59BEAT A RETREAT...
- 4:20THIS WORKFLOW...
- 6:26THE SPREADSHEET
- 9:03A REALLY HARD PROBLEM
- 10:09PHENOMENOLOGY OF THE DATA SCIENTIST
- 11:50CRITERIA
- 12:43PYTHON
- 14:20ASSESSMENT
- 15:40THE COMMON ERRORS
- 17:30THE CONTEMPORARY SCENE
- 17:45THAT IS... • Frame data structure with full API modeled after Pandas
- 19:10SADDLE
- 20:55SCALA NLP
- 21:21MORE RESEARCH
- 22:04SCALA NOTEBOOK
related talks
42mScala: The Unpredicted Lingua Franca for Data Science
47mScala: The Unpredicted Lingua Franca for Data Science
06mInterview
39mMachine Learning with Scala on Spark
54mWhy Netflix ❤'s Scala for Machine Learning
24mQuantum Computing Modeling in Scala
32mCyfra: Scala on GPU
41mSwift + Metal: The future of scientific programming?