PyData London 2018 Lightning Talks - Day 1
summary
PyData lightning compiles short results and concrete tools: NumFOCUS fundraising and community impact, a worldwide pandas sprint that fixes 300 pandas docs fixed, a proposed similarity encoding for high‑cardinality dirty categorical strings, a model‑version study showing surprising jumps (GBM changes), claim‑detection reaching 92% precision 87% recall, a tiny tool mstore leveldb for fast disk-backed word embeddings, practical Python2→3 porting notes, and a PEP‑551 security exploration with poisoned bytecode proofs of concept.
AI-written summary · about
chapters (13)
- 0:00NumFocus mission
- 3:17Open source sustainability
- 7:08Funding and contributions call
- 16:19Dirty categories in ML
- 20:12PyData meetup guide
- 22:21Model version consistency
- 28:51Fact-checkable claims detection
- 33:07Efficient word embeddings loading
- 38:10Python 2 to 3 porting
- 43:26PyLondinium announcement
- 44:52Community appreciation
- 45:56Pub quiz and PEP 551
- 51:10Security in Python interpreters
