Posts for year 2025
- An accessible introduction to mutual information using a d20
- The case for public Slack channels only (no DMs).
- Tornados are just wind at close range.
- Time zones: why you shouldn't use the terms EST or GMT
- The hardest bugs are non-bugs
- Blogging in the LLM age: a second golden age for blogs?
- My review guide for machine learning conference papers
- Problems with ML for toxicity prediction.
- Chebyshev Scalarization Explained
- Reminder that Claude is really good for Chinese grammar
- Some Beijing travel tips
- Taking the V out of VAEs: long live KL-regularized autoencoders!
- Coding python packages with AI
- Why your active learning algorithm may not do better than random
- Generic recommendations for cheminformatics models
- Using LLMs to improve my Chinese
- Conceptual confusion about desirable outputs of reaction prediction models.
- Punishing poor reviewers at CVPR
- Why don't ML conferences provide reviewer instructions?
- Alpha over LLMs
- Is offline model-based optimization a realistic problem? (I'm not convinced)
- Experiment: more posts, lower quality
- Stock responses about statistical significance for reviewing machine learning papers
- Hiring is hard: why good applicants without connections can get overlooked
- Reaction model scores are CRITICAL to multi-step retrosynthesis
- Double checking that Gauche's fingerprint kernels are positive definite
- What ML researchers and users get wrong: optimistic assumptions
- New Year's Resolutions for 2025
- Review of NeurIPS 2024 and predictions for ML in 2025