Toots for cheng@masto.ai account

Written by Cheng Soon Ong on 2025-02-05 at 01:09

Should I sign an agreement that allows my publisher to make deals with other companies to scrape my book for training a large language model?

The book is freely and legally available at: https://mml-book.com

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2025-01-29 at 09:05

My Pebble watch still works, but doesn't talk to my phone any longer. Looking forward to this new one.

https://ericmigi.com/blog/why-were-bringing-pebble-back

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2025-01-21 at 09:44

Arxiv is a pre-print server for academic research.

This is an annual 3 minute anonymous survey about arxiv:

https://cornell.ca1.qualtrics.com/jfe/form/SV_6m22mbqW9GQ3pQO

Please re-post, re-tweet, re-skeet, re-toot, or email it to your colleagues

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2025-01-15 at 09:10

It turns out that even if an adversary introduces some "bad" sentences before showing it to the language generator, the language generator still will be able to learn how to generate good sentences.

https://arxiv.org/abs/2501.04179

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2025-01-15 at 00:07

If you care about theoretical computer science, you should watch this lovely talk by Jon Kleinberg on language generation being easier than language identification. He mentions an interesting tension between hallucination and mode collapse.

https://arxiv.org/abs/2404.06757

https://www.youtube.com/live/BrdaVSZBuyU?feature=shared

[#]MachineLearning #LLM #nlp

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2025-01-07 at 23:18

"... probability probably does not exist — but it is often useful to act as if it does."

David Spiegelhalter provides a short essay that touches on the main aspects of the elusive idea of probability.

⁠https://www.nature.com/articles/d41586-024-04096-5

His book on Uncertainty just came out yesterday, which I expect will explain these ideas in more detail.

⁠https://www.penguin.com.au/books/the-art-of-uncertainty-9780241658628

[#]MachineLearning #Statistics #Probability #scicomm

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2025-01-07 at 02:16

Q: Is machine learning good or bad for the natural science?

A: It depends...

https://proceedings.mlr.press/v235/hogg24a.html

What do you think? #MachineLearning

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2025-01-06 at 10:16

For fans of machine learning theory, Francis Bach's book is available.

https://francisbach.com/my-book-is-out/

[#]MachineLearning

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2024-12-21 at 10:22

[#]Cycling to work is good for you.

This study, based on a large representative sample followed up for 18 years, provides robust evidence of lower risk of all-cause mortality, any hospital admission, cardiovascular disease hospitalisation and medication, cancer incidence and mortality and medication for poor mental health among cycling commuters compared with non-active commuters.

https://bmjpublichealth.bmj.com/content/2/1/e001295

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2024-12-04 at 06:33

I was lucky to be able to catch Bin Yu in person this morning, as she described why truthful data science is so important. She used examples from medical decision making, and spoke about her recent book with Rebecca Barter.

https://vdsbook.com/

[#]MachineLearning #ResponsibleAI

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2024-11-29 at 00:16

What is the chance that: if I repeat an experiment (that has a statistically significant result), the repeated experiment also has a significant result in the same direction?

This paper carefully constructs a relationship between p-values and #replication probability for binary classification.

https://jmlr.org/papers/v25/24-0158.html

[#]MachineLearning #HypothesisTesting

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2024-11-25 at 00:49

Conformal prediction was invented by Vladimir Vovk, more than two decades ago.

https://vovk.net/

It has recently seen a revival in the machine learning community, and there was a keynote by Emanuel Candes in NeurIPS 2022.

https://neurips.cc/virtual/2022/invited-talk/55872

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2024-11-25 at 00:46

A new book on conformal prediction. Conformal prediction is a way to estimate uncertainty of a predictive model, which reminds me a lot of cross validation methods. #MachineLearning

https://arxiv.org/abs/2411.11824

Anastasios N. Angelopoulos, Rina Foygel Barber, Stephen Bates

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2024-11-19 at 23:57

New book on Bayesian inference and human cognition. I have always enjoyed material from Tom Griffiths and also from Josh Tenenbaum, and I expect this new collected chapters would also be excellent. If you want to explore more literature, the contributing authors of individual chapters are also wonderful.

https://mitpress.ublish.com/ebook/bayesian-models-of-cognition-reverse-engineering-the-mind-preview/12799/

[#]MachineLearning #Cognition #bayesian

=> More informations about this toot | View the thread

Written by Cheng Soon Ong on 2024-09-16 at 04:53

There are grants to apply for and prizes to win at the AI in Science Conference, at the ANU in Canberra, in partnership with the Australian Academy of Science. This event is targetted at Early and Mid Career Researchers, and deadlines are coming up soon.

https://bit.ly/AIinScienceConference

[#]MachineLearning #Science #ArtificialIntelligence #AI #Biology #Chemistry #Physics #Astronomy #Psychology

=> More informations about this toot | View the thread

=> This profile with reblog | Go to cheng@masto.ai account

Proxy Information
Original URL
gemini://mastogem.picasoft.net/profile/109396033723075387
Status Code
Success (20)
Meta
text/gemini
Capsule Response Time
384.842422 milliseconds
Gemini-to-HTML Time
3.017285 milliseconds

This content has been proxied by September (3851b).