Our new #ICLR2024 paper shows how LLMs can successfully check their own change of thought reasoning without any fine-tuning or even examples, using an approach we call SelfCheck. Join me at poster 125 this afternoon to learn more Paper: openreview.net/forum?id=pTHf…

We are delighted to announce an ACM-TOPML special issue on "Probabilistic Programming". Please see the attached call for papers for details

We propose alternative schemes that set the weights by optimizing a predictive objective, leveraging ideas from stacking and PAC-Bayes. They can be implemented as cheap post-processing steps on top of existing inferences and lead to more robust weights & better predictions. (5/5)

For probabilistic programs, one damning failure case is that the SLP weights tend to be overconfident, i.e. overly collapsing onto a single SLP. This means inferences might change drastically based on nominal changes in the data and will lead to sub-optimal predictions. (4/5)

Previous work showed that such programs can be decomposed into a mixture of straight-line programs (SLPs), each corresponding to a particular control flow path. The overall program posterior is then a weighted average over the local posteriors of the SLPs. (2/5)

The typical Bayesian NN setups treats all parameters stochastically. @MrinankSharma will shortly present our @aistats_conf "Notable Paper" on why this fundamental assumption is usually not justified and partially stochastic networks are often preferable proceedings.mlr.press/v206/s…

We've got really good at utilizing data. But methods for acquiring that data are often still rudimentary. Our new review paper shows how Bayesian experimental design has recently transformed to now provide a powerful mechanism to acquire data intelligently arxiv.org/abs/2302.14545

Interested in the optimal design of experiments, adaptive data acquisition, or active learning? Then check out my tutorial on Bayesian Experimental Design from the Cambridge Ellis Machine Learning Summer School Part 1: youtu.be/5oGdDGPVLjc Part 2: youtu.be/yCHp_wzpq8o

Interesting in performing variational inference and parameter learning in a fully online manner? The poster session for our @NeurIPSConf oral Online Variational Filtering and Parameter Learning is starting now at neurips.cc/virtual/2021/post… Paper: openreview.net/forum?id=et2s…

Interested in adaptive experiments that can run in real-time? ⏱️🧪 Then come join our @NeurIPSConf poster session for Implicit Deep Adaptive Design: Policy-Based Experimental Design without Likelihoods Happening *now*: neurips.cc/virtual/2021/post… Paper: openreview.net/forum?id=on2D…