Simple Machine Learning
summary
Claims a working classifier that labels Join In events using PHP code called enamel and multivariate and Bernoulli naive Bayes. Data: 1,084 events, 409 tagged, 675 untagged; original estimate of new speakers ~49% drops to ~44% after classifier filtering. Model build takes tens of seconds, prediction ~30 ms per event. Implementation hits memory limits with ~350 tags and ~13,000 features (~300+ MB in PHP arrays) so feature selection is needed.
AI-written summary · about
chapters (23)
- 0:00Intro
- 0:18Who is this guy?
- 1:14The Challenge
- 1:55How Insular is PHP?
- 4:20Solutions
- 5:29Mechanical Turk
- 7:26Pattern Recognition
- 10:02Teaching
- 12:18Unsupervised Learning
- 15:37Regression
- 16:54Classification
- 19:39Training
- 19:54Feature Extraction
- 23:20Feature Vectors
- 27:21Probabilistic Correlation
- 30:21Bayes' Theorem
- 31:14Multivariate Bayes'
- 34:44Demo Time
- 42:16Human Learning
- 42:30Lessons Learned
- 45:18Next Steps
- 48:05The Code
- 48:45Questions?
