Agile Experiments in Machine Learning with F#

Explore the world of machine learning through the lens of agile experimentation in this 58-minute conference talk. Dive into the unique workflow of ML projects, contrasting it with traditional software development. Learn how to tackle challenges in ML competitions, using the Kaggle Home Depot competition as a case study. Discover techniques for setting up an efficient experimental harness, maintaining code sanity, and adapting software development principles to the ML context. Gain insights into the iterative nature of ML model development, the importance of rapid experimentation, and strategies for validating progress in this dynamic field.

Intro
About me
Code on GitHub
Kaggle Home Depot
What we want
What we really mean
Overall process
ML: experiments in code
Learning with Algorithms
What are the problems?
Core model
Catalog of Features
Did it work?
Dataset normalization
Pre-processing pipeline
Lesson learnt
Tension
General