Machine learning in your hobby project: Obstacles, entry points and solutions
summary
Personal bookmarking app holds about 1500 links, 74 categories and shows slow, painful category selection so a simple classifier is added to suggest picks. Implementation uses a handrolled Naive Bayes with smoothing over tokenized URLs and titles, yielding candidate probabilities around 1e-18 to 1e-24 but still ranking correctly in the UI. Training stays lightweight and training on the fly for now; next steps are stop‑word handling, n‑grams, lemmatization, SVMs and clustering.
AI-written summary · about
chapters (7)
- 0:00Start
- 8:00Hobby project: Inception
- 10:03Hobby project: some domain concepts
- 21:06What to do with the data?
- 26:33Bayes' Rule: Example (cont.)
- 30:29Bayes' Rule: Why all the small values? ?
- 33:43Next steps
related talks
36mMachine Learning Microservices in Rails
48mHow to make intelligent web-apps
39mMachine learning
39mHow to Avoid structural biases in evaluating your Machine Learning/NLP projects
21mMachine Learning and Cyber Security - Detecting malicious URLs in the haystack
35mIntelligent Ruby: Getting Started with Machine Learning
52mMachine Learning Crash Course
31mWhen humans teach machines