FP for Data Science For Loops Considered Harmful

on YouTube ↗21m

" summary

Argues that using static types and immutability prevents whole classes of bugs (matrix-dimension and unit mismatches become compiler errors), replaces brittle for-loops with map and fold for scale-invariant code that runs the same on a laptop or a Spark cluster, and models effects as ADTs/monads so functions are total and testable. Shows Python/Scala examples, sketches a free monad DSL that swaps interpreters for scikit-learn, TensorFlow or Spark.

AI-written summary · about

" chapters (12)

1. 0:00Introduction and motivation
2. 2:06Audience engagement questions
3. 3:50Key FP concepts explained
4. 7:47Benefits of FP for data science
5. 10:18Python vs functional trade-offs
6. 12:05Parallelism in data pipelines
7. 13:07Correctness through FP
8. 15:16Expressiveness in functions
9. 16:23Algebraic properties in code
10. 17:07Compositionality in ML
11. 18:32FP in neural networks
12. 19:35Call to embrace FP

" related talks

• 40mFunctional Programming for Machine Learning and Data PipelinesMalcolm Greaves · Data By the Bay 2016
• 01h 44mFunctional Programming: Technical Reasons to AdaptVenkat Subramaniam · JavaDay Kyiv 2015
• 45mFP: The Good, the Bad and the UglyDaniela Sfregola · ScalaIO 2019
• 47mKeeping the fun in Apache Spark Datasets and FPHolden Karau · Scala Days Berlin 2018
• 01h 00mWhat The Functional ProgrammingJames Ward · Great International Developer Summit 2025
• 27mFunctional Programming for PythonistasBianca Gibson · PyCon AU 2016
• 59mFunctional Programming for Machine Learning PanelVitaly Gordon, Malcolm Greaves · Scale By the Bay 2017
• 44mFunctional Programming You Already KnowKevlin Henney · Codemotion Rome 2015