Joel Dyer

There can be no better example of something deeply wrong with this country's national culture than the normalisation of the description of a soggy pre-packaged sandwich as a "main".

Check out the paper below, or come chat to us in Vancouver for more details!

Paper: lnkd.in/eHjtXYbx

#AI4Science #Bayes #ICML #simulation

We propose and compare two ways to learn RPs for arbitrary simulation models using neural mutual information estimators, showing that these approaches can accurately recover known reference priors and provide a way to subsequently perform neural simulation-based Bayesian inference *for free*.

Finding reference priors is difficult, however – this is especially true for simulation models, which are used across scientific domains (particle physics simulators in physics, agent-based models in economics, etc.).

This is measured by a KL divergence between the posterior and prior. The idea is that these priors provide a means to conduct prior sensitivity analyses, allowing modellers to measure how much prior information is incorporated into their Bayesian procedures through the prior distribution they chose.

RPs were developed as one possible way to formalise the notion of “minimal prior information” in Bayesian inference, and are priors that maximise the expected information to be gained from data about the model’s parameters.

We’ve got a new paper out on “Learning Likelihood-free Reference Priors”, forthcoming at #ICML2025.

In this paper, we consider how reference priors (RPs) can be constructed for arbitrary simulation models. 🧵

CAR Causal Abstractions and RepresentationsWorkshop @ UAI 2025July 25th 2025, Rio de Janeiro 🇧🇷

🔥 Got a great work on causal representation learning, abstraction, high-dimensional discovery, or other hot topics in causality?

🇧🇷 Don’t miss your chance to present in Rio at the CAR Workshop at #UAI2025!

⏰ Deadline is in 1 week – May 26!
🌐 sites.google.com/view/car-25/

I got fooled by AI-for-science hype—here's what it taught me I used AI in my plasma physics research and it didn’t go the way I expected.

Going to the hospital because I broke my wrist smashing the endorse button:
www.understandingai.org/p/i-got-fool...

How the web became unreadable, an article obscured by medium's pop up asking you to sign up to read the words that have already loaded.

Lmao

1. LLM-generated code tries to run code from online software packages. Which is normal but
2. The packages don’t exist. Which would normally cause an error but
3. Nefarious people have made malware under the package names that LLMs make up most often. So
4. Now the LLM code points to malware.

AI teddies could ‘harvest data from your children’ Sir Nigel Shadbolt, a leading researcher, has raised concerns about ‘smart toys’

Professor Nigel Shadbolt has shared his thoughts on the potential risks of AI-powered teddy bears that use AI to converse with children. He suggested that they 'could be used to survey, they could be used to appropriate and harvest data from the child.' www.telegraph.co.uk/news/2025/03...

Mencap campaigners posing next to posters in the Westminster tube station. The text says "People with a learning disability die, on average, up to 20 years younger than the general population, even though this could be avoided with the right healthcare."

Mencap campaigners posing next to posters in the Westminster tube station. The text says "People with a learning disability die, on average, up to 20 years younger than the general population, even though this could be avoided with the right healthcare."

1/1🗣️People with a learning disability die up to 20 years younger than the general population, even though this could be avoided with the right healthcare.

Agent Based Modelling Comes of Age Our recently published research demonstrates how ABMs is 'coming of age' cross three areas: central banking, economics and finance, and AI.

🆕 Agent Based Modelling Comes of Age
@doynefarmer.bsky.social believes creating economic models that can make useful predictions may be the most important problem in economics today
🔖We summarise 3 of 2025's ground breaking papers
🔗https://www.inet.ox.ac.uk/news/agent-based-modelling-comes-of-age

Excited to announce we are organising a two-day workshop on 'Bayesian Computation and Inference in Misspecified Models' at BayesComp 2025!

Registrations are now open, with a very nice lineup of speakers to be announced soon! bayescomp2025.sg

@isba-bayesian.bsky.social

We’re overwhelmed by all the generous donations from Traitors fans after hearing Alexander Dragonetti’s story! 🎉

You’d wanted to know how much has been raised...

#TheTraitors #AlexanderDragonetti

Definitely consider submitting your work to UAI, it's fantastic conference!

And while you're at it, why not submit a shorter version to the AABI workshop track (deadline Feb 7)? 😉

#UAI is one of the best general conferences for causal researchers

Submission deadline: Feb 10

#CausalSky

✨I made a starter pack for complexity scientists on this platform.✨

go.bsky.app/QzMwdrL

Next week will be ✈️to @neuripsconf.bsky.social together with @joelnmdyer.bsky.social and Nick Bishop to present our work on "Interventionally Consistent Surrogates for Complex Simulation Models". Details 👇

@neuripsconf.bsky.social

@inetoxford.bsky.social @inet-complexity.bsky.social @compscioxford.bsky.social #causal #simulation #Bayes

2 weeks left to apply for this PhD position (link at end of thread) #EpiSky #IDSky

2) “Model Exploration through Marginal Likelihood Entropy Maximisation”, accepted at the #NeurIPS2024 D3S3 Workshop. We propose a method for exploring the behaviours a (differentiable) simulator can generate – useful for understanding your mechanistic model! openreview.net/pdf?id=dPszK... (3/3)

This provides decision- and policy-makers with a way to reduce the cost of running simulations when designing policies and interventions. Useful in time-critical settings, such as emerging pandemics! (2/3)

See you at #NeurIPS2024! We’ll present two papers:

1) “Interventionally Consistent Surrogates for Complex Simulation Models”. Here, we propose a method for learning causal abstractions (“interventionally consistent surrogates”) of complex simulation models! openreview.net/pdf?id=UtTjg... (1/3)

No. 2024-06 - Forecasting Macroeconomic Dynamics using a Calibrated… In the last few years, economic agent-based models have made the transition from qualitative models calibrated to match stylised facts to quantitative…

New paper on Forecasting with Agent-Based Models from Sam Wiese, Jagoda Kaszowska-Mojsa, @joelnmdyer.bsky.social, @j0mrn.bsky.social, Marco Pangallo, Francois Lafond, John Muellbauer, Ani Calinescu, @doynefarmer.bsky.social
#Complexity #EconSky

www.inet.ox.ac.uk/publications...

Researcher wins best paper award for work on approximate Bayesian… It seeks to aid widespread adoption of simulators in real-world modelling and decision-making scenarios, by making approximate Bayesian inference more…

Congratulations to @joelnmdyer.bsky.social for winning best paper at the 40th Conference on Uncertainty in Artificial Intelligence, for his work on approximate Bayesian inference methods for simulation models.
www.inet.ox.ac.uk/news/researc...

blackbirds Black-Box Inference foR Differentiable Simulators.

Our BlackBIRDS python package implements inference methods for differentiable simulators in PyTorch, and features heavily in the tutorial. You can play with it yourself by installing from pypi.org/project/blac..., and check out an example use case at github.com/joelnmdyer/g....

Our group recently delivered a tutorial on building & calibrating differentiable agent-based models of financial systems at the 4th International Conference on AI in Finance #ICAIF2024!

Notebooks & materials are freely available on our tutorial website: largeagentcollider.github.io/icaif_tutori....