Using feature importance to interpret your models? This paper might be of interest to you. Papers by @gunnark.bsky.social are always worth checking out.

Gunnar König@gunnark.bsky.social · last yr.

In many XAI applications, it is crucial to determine whether features contribute individually or only when combined. However, existing methods fail to reveal cooperations since they entangle individual contributions with those made via interactions and dependencies. We show how to disentangle them!

My stock portfolio is deep in the red, and tariffs by the Trump admin might be the cause. Could an LLM have been used to calculate them? It made me rethink how LLMs shape decisions, from big global-economy-wrecking ones to everyday decisions.

Who’s Really Making the Decisions?

LLMs, tariffs, and the silent takeover of decisions

mindfulmodeler.substack.com

SHAP interpretations depend on background data — change the data, change the explanation. A critical but often overlooked issue in model interpretability. Read more:

SHAP Interpretations Depend on Background Data — Here’s Why

Or why height doesn't matter in the NBA

mindfulmodeler.substack.com

I recently joined The AI Fundamentalists with my co-author Timo Freiesleben to discuss our book Supervised Machine Learning for Science. We explored how scientists can leverage ML while maintaining rigor and embedding domain knowledge.

Supervised machine learning for science with Christoph Molnar and Timo Freiesleben, Part 1 - The AI Fundamentalists

Machine learning is transforming scientific research across disciplines, but many scientists remain skeptical about using approaches that focus on prediction over causal understanding. That’s why...

buzzsprout.com

The 3rd edition of Interpretable Machine Learning is out! 🎉 Major cleanup, better examples, and new chapters on Data & Models, Interpretability Goals, Ceteris Paribus, and LOFO Importance. The book remains free to read for everyone. But you can also buy ebook or paperback.

Has anyone seen Counterfactual Explanations for machine learning models somewhere in the wild? They are often discussed in research papers, but I have yet to see them being used somewhere in an actual process or product.

Trying Claude Code for some tasks. Paradoxically, it's most expensive when it doesn't work because it fails, then tries a couple of times again, burning through tokens. So sometimes it's 20 cents for saving you 20 minutes of work. Other times it's $1 for wasting 10 minutes.

Only waiting for the print proof, but if it looks good, I'll publish the third edition of Interpretable Machine Learning next week. As always, it was more work than anticipated—especially moving the entire book project from Bookdown to Quarto, which took a bit of effort.

Can an office game outperform machine learning? My most recent post on Mindful Modeler dives into the wisdom of the crowds and prediction markets. Read the full story here:

Can an office game outperform machine learning?

Wisdom of the crowds, prediction markets, and more fun in the work place.

buff.ly

A year ago, I took a risk & spent quite some time on a ML competition. It paid off—I won 4th place overall & 1st in explainability! Here's a summary of the journey, challenges, & key insights from my winning solution (water supply forecasting)

OpenAI right now

The original SHAP paper has been cited over 30k times. The paper showed that attribution methods, like LIME and LRP, compute Shapley values (with some adaptations). The paper also introduces estimation methods for Shapley values, like KernelSHAP, which today is deprecated.

To this day, the Interpretable Machine Learning book is still my most impactful project. But as time went on, I dreaded working on it. Fortunately, I found the motivation again and I'm working on the 3rd edition. 😁 Read more here:

Why I almost stopped working on Interpretable Machine Learning

7 years ago I started writing the book Interpretable Machine Learning.

buff.ly

How I sometimes feel working on "traditional" machine learning topics instead of generative AI stuff 😂

It's quite ironic how people who built the best prediction models are such bad predictors themselves. They throw all their knowledge about how to make good predictions overboard and just claim things like: AI will replace radiologists in a few years or when they expect AGI.

The problem with all these AI demos (especially for image and video generation): They are the most impressive, cherry-picked examples. That includes cherry-picking prompts and themes that produced better results. But as a user, you want good results for every prompt/theme relevant to your use case

Looking for a Christmas gift for a stubborn Bayesian or an over-hyped AI enthusiast? Modeling Mindsets is a short read to broaden your perspective on data modeling. christophmolnar.com/books/modeli... *Hat not included.

My personal rules for AI-assisted writing: • Use AI only for small and specific stuff, like grammar fixes or making suggestions for factual corrections. • Never let an LLM change voice and tone. • I review any changes made by AI.

What a sad timeline, where vaccines — one of medicine's clearest wins with all upside and minimal downside — have become targets. Can't we have like an anti-knee arthroscopy movement or whatever instead?

Citing a non-deterministic, "hallucinating", and non-reproducible LLM output is wild. While the norms and best practices are evolving, citing them seems the wrong way. (even wilder when some people add "ChatGPT" as their co-authors)

Arvind Narayanan@randomwalker.bsky.social · 2y ago

One of the weirder scholarly practices regarding generative AI that seems to have been normalized is citing chatbots. I say normalized because many univs & scholarly associations recommend it as an element of proper scholarship. But it doesn't make sense when you consider what a citation means. 1/

What are shapley interactions and why should you care about them? This is a guest post by Julia, Max, Fabian and Hubert on my newsletter Mindful Modeler. I also learned a lot from this post and definitely recommend checking out the shapiq package. mindfulmodeler.substack.com/p/what-are-s...

What Are Shapley Interactions, and Why Should You Care?

A guest post by Julia, Max, Fabian and Hubert.

mindfulmodeler.substack.com

Is anyone aware of a completely AI-generated book that people actually read? Excluding books that are dedicated "AI experiments" and where the book is more about the experiment. Also excluding AI-assisted books where generative AI played a minor role

The unofficial GIF-based pandas library documentation. pandas.DataFrame.rolling

a panda bear is laying down in the grass .

Alt: a panda bear is rolling down in the grass. It's a side-ways roll, hlding some type of object. I give 10/10.

media.tenor.com

Without non-linear activation functions, neural networks would be linear models, no matter how many layers are stacked.

Santiago@svpino.com · 2y ago

Without non-linear activation functions, neural networks would be too weak to solve anything interesting. Here is a simple way to understand how powerful non-linearities are: 1/7

Got myself a Samsung Galaxy S9 tablet for note-taking, and I love it. (and yes, I'm reading my own book here as a reference for another project, feeling like an imposter because I don't have everything memorized 😂)

I'm always amazed at how popular the random forest algorithm is for remote sensing research. I would have expected deep learning to be more popular there (not to say it isn't). Must be an attraction to trees. 😁

What's the difference between explainability and interpretability? Does the machine learning community have an agreed-upon definition? No. There's a great overview, which is from this paper: arxiv.org/abs/2211.08943 My take: I prefer interpretability since the term explainability is too strong.

AI for science could be more impactful than chatbots. It is already helping win Nobel prizes and accelerating drug development and materials discovery. Today we published an essay about it: why it matters, how it’s happening and its implications. Here is a summary from an econ / social sci lens.

Machine learning and statistics have very narrow ideas of what a model is and how to abstract the world. To broaden the "model" horizon, I can recommend these two books: • Thinking in Systems by Donella H. Meadows • Simulation and Similarity by Michael Weisberg (ignore the monkey)

No one can explain stochastic gradient descent better than this panda.

a panda bear is rolling around in the grass in a zoo enclosure .

Alt: a panda bear is rolling around in the grass in a zoo enclosure .

media.tenor.com