Michael Lones

A mutualistic collection of cells. Also a @HeriotWattUni academic doing research into machine learning, bio-inspired computing and other interesting things.

[Michael Lones @michael_lones]

I don’t hang out here anymore. You can find me at https://t.co/8nk96P1eaC or @michaellones.bsky.social

[Michael Lones @michael_lones]

Great to have been involved in this initiative led by @sayashk and @random_walker to (hopefully!) improve the use of machine learning in science. Further thoughts in my Substack post: https://t.co/pgCDOqWWdC

[Sayash Kapoor @sayashk]

Excited to share that our paper introducing the REFORMS checklist is now out @ScienceAdvances!

In it, we:
- review common errors in ML for science
- create a checklist of 32 items applicable across disciplines
- provide in-depth guidelines for each item

https://t.co/M21VWGIZhC

[Michael Lones @michael_lones]

I’ve been meaning to write a book about computers for years. However, this requires time that I probably don’t have. Instead, I’ve decided to set up a substack: https://t.co/z0Sdjnnrz9

[Kahraman Koştaş @kkostas]

📄New Paper Alert! 🌐 "IoTGeM: Generalizable Models for Behaviour-Based IoT Attack Detection" with @justmikejust @michael_lones.🚀This paper tackles the challenge of adapting IoT attack detection models to new data with an innovative approach.
https://t.co/CPDT97CFQJ

[Michael Lones @michael_lones]

New year, new pitfalls! I’ve just updated “How to avoid machine learning pitfalls: a guide for academic researchers” to v4, with new material on LLMs, ensembles, checklists, and data usage, and numerous tweaks elsewhere. https://t.co/CKjUgyE9A9

[Sayash Kapoor @sayashk]

ML-based science is facing a reproducibility crisis. We think clear reporting standards for researchers can help.

Today, we're introducing REFORMS, a consensus-based checklist authored by 19 researchers across many disciplines.
https://t.co/p9ISyyRnV0

[Michael Lones @michael_lones]

@justmikejust @FlorenceHRScott Why compromise? https://t.co/HCbUXw2rIZ

[Michael Lones @michael_lones]

I wonder if it was a conscious decision not to include the word “Answering” in “Prime Minister’s Questions”?

[Michael Lones @michael_lones]

Would be nice if Google Scholar could sort by interestingness, perhaps as decided by a LLM. My most interesting papers (as decided by me) are buried deep in my cited-by list.

[Michael Lones @michael_lones]

@RestIsPolitics @RoryStewartUK @campbellclaret Who should decide the future of AI, technologists or politicians?

[Michael Lones @michael_lones]

Interesting to hear political bias in “AI chatbots” discussed on #PoliticsLive today, though the implication this is due to left-wing programmers suggests a lack of understanding in political circles about how these things work.

[Michael Lones @michael_lones]

Women in Data Science Edinburgh will be held at Heriot-Watt University (Hybrid) on 15th May 2023 and invites contributions from women who use data science: https://t.co/ASiPxKyrZv

[Michael Lones @michael_lones]

@RestIsPolitics @LondonPalladium @RoryStewartUK @campbellclaret Media coverage of ChatGPT etc has focused on plagiarism. A greater concern is automation of highly-skilled jobs. This paper, just released, lists many of the professions which are in danger: https://t.co/ZfTV7gnUZp. Any thoughts on the social, economic and political consequences?

[Michael Lones @michael_lones]

Updated “How to avoid machine learning pitfalls”. New sections on handling spurious correlations, temporal dependencies in time series data, and keeping up with deep learning, plus lots more references: https://t.co/CKjUgyE9A9

[Michael Lones @michael_lones]

Happy memories…
1.BASIC
2.6502 Assembly
3.Forth
4.C
5.HyperTalk
6.Scheme
7.Ada
8.Haskell
9.Prolog
10.C++
https://t.co/ROeb7eIcod
12.PHP
13.R
14.Python
15.Kotlin

[naomi (forgetful functor) @fixedpointfae]

in what order did you learn your languages?

1. python
2. R
3. haskell
4. C
5. rust
6. scheme
7. {java,type}script
8. clojure
9. scala
10. java
11. idris
12. go

[Michael Lones @michael_lones]

@[email protected]

[Michael Lones @michael_lones]

An interesting paper on why tree-based ensemble models, such as random forest, tend to outperform deep learning models on tabular (i.e. not image or text) data. https://t.co/d3oWzpHDV7

[Michael Lones @michael_lones]

Trying to find a reasonable explanation for a political speech I saw earlier.

[Michael Lones @michael_lones]

Thanks to everyone who gave feedback on version 1. Please keep the feedback coming!

[Michael Lones @michael_lones]

Just updated "How to avoid machine learning pitfalls” to version 2. New sections on data augmentation pitfalls and when not to use deep learning, plus more on leakage, stats and transparency, new references and various tweaks. https://t.co/CKjUgyE9A9