5 notes · 5 topics
Notes
Papers, articles, podcasts, reports and talks, each reduced to the part worth keeping. Notes link to each other with [[wikilinks]]; thegraph is the same set seen sideways.
Topic
Source
Year
- How notes work hereThe conventions this knowledge base runs on — frontmatter, wikilinks, topics and maturity. Read this one first.Generative VisualsSkiing
- Attention is all you needThe Transformer paper. Drops recurrence entirely, and in doing so makes sequence modelling a parallel problem rather than a sequential one.Neural Networks
- Loss landscapes and terrainSeed. The slope-reading intuition from backcountry skiing keeps wanting to be a statement about optimisation geometry, and I have not made it rigorous yet.Neural NetworksSkiing
- Terrain as a first filterSeed. Terrain choice does more risk reduction than snowpack assessment, because it is the variable you actually control.Snow ScienceSkiing
- Sequence and sight-readingSeed. Reading a boulder problem from the ground is planning under a model you know to be wrong; the skill is in how cheaply you revise.Climbing