iiinova — where your digital self keeps evolving

Tommy Rochussen @rochussen.bsky.social · 22/04/2026

If you're at ICLR and want to chat about amortised inference/neural processes or probabilistic ML more broadly, swing by my poster between 10:30-13:00 on Thursday (poster session 1) in Pavilion 3 at location #206. The poster is about NP-style learning in BNNs to find good priors.

041

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

Check out the paper 👉 arxiv.org/pdf/2602.087... Looking forward to presenting this work in Rio, and many thanks to @vincefort.bsky.social for his supervision!

arxiv.org

031

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

Are humble Gaussian priors enough for BNNs to model highly complex stochastic processes? Do well-specified BNN priors remove the need for more costly approximate inference algorithms? We provide answers in the paper!

110

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

2. It turns BNNs into flexible generative models (i.e., sampling from learned priors. 3. It enables capabilities that have been difficult for neural processes so far, including: • Within-task minibatching • Meta-learning in extremely data-scarce regimes.

110

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

Why this matters: 1. It lets us study BNNs under well-specified, data-driven priors rather than the usual isotropic guff.

100

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

3. The resulting model can be viewed as a neural process whose latent variable is the weights of a BNN, with the network itself acting as the decoder.

100

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

2. This is achieved via per-dataset amortised variational inference, allowing the model to infer dataset-specific posteriors while learning a shared, well-specified prior.

100

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

What we do: 1. We propose a way to learn a prior over neural network weights from data, using a collection of related datasets.

100

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

Bayesian neural network (BNN) practitioners have to specify priors over weights, but doing so is often unclear or ad hoc. In this paper, we bridge Bayesian deep learning and probabilistic meta-learning to offer a concrete answer.

120

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

The work tackles a fairly fundamental question in Bayesian deep learning: "how can we be Bayesian if we don’t have any meaningful prior beliefs in the first place?"

100

Tommy Rochussen @rochussen.bsky.social · 11/02/2026

I’m pleased to share that our latest paper, “Amortising Inference and Meta-Learning Priors in Neural Networks”, has been accepted to ICLR 2026 in Rio!

131

Tommy Rochussen @rochussen.bsky.social · 24/01/2026

Are bitterns as fiendishly difficult to spot in Singapore as they are in Europe?

110

Tommy Rochussen @rochussen.bsky.social · 17/04/2025

Arxiv link: arxiv.org/pdf/2504.01650 It’s nice to be able to get the ball rolling on my PhD with this paper, and a nice achievement to have published my first non-workshop paper. A big thanks to @vincefort.bsky.social for his supervision on this project!

arxiv.org

010

Tommy Rochussen @rochussen.bsky.social · 17/04/2025

1.) you want/need GP levels of interpretability 2.) you don’t have that many training tasks, so need SOTA data efficiency (at the meta-level) 3.) you have accurate domain knowledge (in GP-prior form) 4.) each task has too many observations for exact GP inference

110

Tommy Rochussen @rochussen.bsky.social · 17/04/2025

If you need probabilistic predictions across multiple related tasks/datasets, you should use this model if any combination of the following hold:

110

Tommy Rochussen @rochussen.bsky.social · 17/04/2025

We introduce the ability to meta-learn sparse variational Gaussian process inference, resulting in a new type of neural process that is amenable to prior elicitation.

110

Tommy Rochussen @rochussen.bsky.social · 17/04/2025

Very pleased to share that our new paper “Sparse Gaussian Neural Processes” has been accepted under the proceedings track at AABI 2025! 🎉 (1/n)

270

Reposted by Tommy Rochussen

antonio vergari ⚔️ short-circuiting @nolovedeeplearning.bsky.social · 26/11/2024

I've seen things you people wouldn't believe. Attacks from reviewers on fire off the shoulders of #OpenReview. I watched logic fallacies glitter in the dark near @iclr-conf.bsky.social All those moments will be lost in time, like tears in the next resubmission.  Time to die. #ML #Ai #PhDlife

2163

Tommy Rochussen @rochussen.bsky.social · 20/11/2024

Thanks for putting this together - keen to be added!

100