no code implementations • 29 Mar 2024 • Mauro Comi, Alessio Tonioni, Max Yang, Jonathan Tremblay, Valts Blukis, Yijiong Lin, Nathan F. Lepora, Laurence Aitchison
Touch and vision go hand in hand, mutually enhancing our ability to understand the world.
no code implementations • 29 Feb 2024 • Xi Wang, Laurence Aitchison
We propose a batch size invariant version of Adam, for use in large-scale, distributed settings, in which the mini-batch is divided into micro-batches which are distributed among worker nodes.
no code implementations • 20 Feb 2024 • Adam X. Yang, Maxime Robeyns, Thomas Coste, Jun Wang, Haitham Bou-Ammar, Laurence Aitchison
To ensure that large language model (LLM) responses are helpful and non-toxic, we usually fine-tune a reward model on human preference data.
no code implementations • 9 Feb 2024 • Ben Anson, Edward Milsom, Laurence Aitchison
A common theoretical approach to understanding neural networks is to take an infinite-width limit, at which point the outputs become Gaussian process (GP) distributed.
no code implementations • 1 Feb 2024 • Theodore Papamarkou, Maria Skoularidou, Konstantina Palla, Laurence Aitchison, Julyan Arbel, David Dunson, Maurizio Filippone, Vincent Fortuin, Philipp Hennig, Jose Miguel Hernandez Lobato, Aliaksandr Hubin, Alexander Immer, Theofanis Karaletsos, Mohammad Emtiyaz Khan, Agustinus Kristiadi, Yingzhen Li, Stephan Mandt, Christopher Nemeth, Michael A. Osborne, Tim G. J. Rudner, David Rügamer, Yee Whye Teh, Max Welling, Andrew Gordon Wilson, Ruqi Zhang
In the current landscape of deep learning research, there is a predominant emphasis on achieving high predictive accuracy in supervised tasks involving large image and language datasets.
no code implementations • 21 Nov 2023 • Mauro Comi, Yijiong Lin, Alex Church, Alessio Tonioni, Laurence Aitchison, Nathan F. Lepora
To address these challenges, we propose TouchSDF, a Deep Learning approach for tactile 3D shape reconstruction that leverages the rich information provided by a vision-based tactile sensor and the expressivity of the implicit neural representation DeepSDF.
no code implementations • 29 Sep 2023 • Xi Wang, Laurence Aitchison, Maja Rudolph
To address these issues, we propose an ensemble approach using Low-Rank Adapters (LoRA), a parameter-efficient fine-tuning technique.
1 code implementation • 18 Sep 2023 • Edward Milsom, Ben Anson, Laurence Aitchison
Recent work (A theory of representation learning gives a deep generalisation of kernel methods, Yang et al. 2023) modified the Neural Network Gaussian Process (NNGP) limit of Bayesian neural networks so that representation learning is retained.
1 code implementation • 6 Sep 2023 • James Malkin, Cian O'Donnell, Conor Houghton, Laurence Aitchison
The resulting networks revealed a tradeoff between circuit performance and the energetic cost of synaptic reliability.
2 code implementations • 24 Aug 2023 • Adam X. Yang, Maxime Robeyns, Xi Wang, Laurence Aitchison
Low-rank adaptation (LoRA) has emerged as a new paradigm for cost-efficient fine-tuning of large language models (LLMs).
no code implementations • 23 May 2023 • Sebastian Ober, Ben Anson, Edward Milsom, Laurence Aitchison
When the distribution is chosen to be Wishart, the model is called a deep Wishart process (DWP).
no code implementations • 18 May 2023 • Thomas Heap, Gavin Leech, Laurence Aitchison
Attaining such a large number of importance samples is intractable in all but the smallest models.
no code implementations • 22 Feb 2023 • Adam X. Yang, Laurence Aitchison, Henry B. Moss
In Bayesian optimisation, we often seek to minimise the black-box objective functions that arise in real-world physical systems.
no code implementations • 8 Feb 2023 • Hugh Panton, Gavin Leech, Laurence Aitchison
These perform well in the presence of complex interactions, with tree depth governing the order of interactions.
no code implementations • 2 Feb 2023 • Jack R. P. Hanslope, Laurence Aitchison
RL is increasingly being used to control robotic systems that interact closely with humans.
no code implementations • 29 Nov 2022 • Henry Addison, Elizabeth Kendon, Suman Ravuri, Laurence Aitchison, Peter AG Watson
This work demonstrates the effectiveness of diffusion models, a form of deep generative models, for generating much more cheaply realistic high resolution rainfall samples for the UK conditioned on data from a low resolution simulation.
1 code implementation • 15 Sep 2022 • Frederik Benzing, Simon Schug, Robert Meier, Johannes von Oswald, Yassir Akram, Nicolas Zucchet, Laurence Aitchison, Angelika Steger
Neural networks trained with stochastic gradient descent (SGD) starting from different random initialisations typically find functionally very similar solutions, raising the question of whether there are meaningful differences between different SGD solutions.
no code implementations • 23 Aug 2022 • Michele Garibbo, Casimir Ludwig, Nathan Lepora, Laurence Aitchison
We therefore tested three major families of modern deep RL algorithm on a mirror reversal perturbation.
1 code implementation • 24 Jun 2022 • Xi Wang, Laurence Aitchison
We develop ShiftMatch, a new training-data-dependent likelihood for robustness to corruption in Bayesian neural networks (BNNs).
no code implementations • 8 Sep 2021 • Anupam K. Gupta, Laurence Aitchison, Nathan F. Lepora
In addition, the unsheared tactile images give a faithful reconstruction of the contact geometry that is not possible from the sheared data, and robust estimation of the contact pose that can be used for servo control sliding around various 2D shapes.
no code implementations • 30 Aug 2021 • Adam X. Yang, Maxime Robeyns, Edward Milsom, Ben Anson, Nandi Schoots, Laurence Aitchison
In particular, we show that Deep Gaussian processes (DGPs) in the Bayesian representation learning limit have exactly multivariate Gaussian posteriors, and the posterior covariances can be obtained by optimizing an interpretable objective combining a log-likelihood to improve performance with a series of KL-divergences which keep the posteriors close to the prior.
1 code implementation • NeurIPS 2021 • Sebastian W. Ober, Laurence Aitchison
We develop a doubly-stochastic inducing-point inference scheme for the DWP and show experimentally that inference in the DWP can improve performance over doing inference in a DGP with the equivalent prior.
no code implementations • 6 Jul 2021 • Laurence Aitchison, Stoil Ganev
In particular, we show that in the infinite sample limit, and for a particular choice of prior, the actual InfoNCE objective is equal to the ELBO (up to a constant); and the ELBO is equal to the marginal likelihood with a deterministic recognition model.
no code implementations • 10 Jun 2021 • Seth Nabarro, Stoil Ganev, Adrià Garriga-Alonso, Vincent Fortuin, Mark van der Wilk, Laurence Aitchison
Here, we provide several approaches to developing principled Bayesian neural networks incorporating data augmentation.
1 code implementation • 14 May 2021 • Vincent Fortuin, Adrià Garriga-Alonso, Mark van der Wilk, Laurence Aitchison
Bayesian neural networks have shown great promise in many applications where calibrated uncertainty estimates are crucial and can often also lead to a higher predictive performance.
no code implementations • 27 Feb 2021 • Ali Unlu, Laurence Aitchison
We develop variational Laplace for Bayesian neural networks (BNNs) which exploits a local approximation of the curvature of the likelihood to estimate the ELBO without the need for stochastic sampling of the neural-network weights.
no code implementations • pproximateinference AABI Symposium 2022 • Xi Wang, Laurence Aitchison
In particular, aleatoric uncertainty signals a specific type of OOD point: one without a well-defined class-label, and our model of data curation gives a likelihood for these points, giving us a mechanism for conditioning on outlier points and thus performing principled Bayesian outlier exposure.
Out-of-Distribution Detection Out of Distribution (OOD) Detection
1 code implementation • NeurIPS Workshop ICBINB 2020 • Vincent Fortuin, Adrià Garriga-Alonso, Sebastian W. Ober, Florian Wenzel, Gunnar Rätsch, Richard E. Turner, Mark van der Wilk, Laurence Aitchison
Isotropic Gaussian priors are the de facto standard for modern Bayesian neural network inference.
no code implementations • pproximateinference AABI Symposium 2021 • Ali Unlu, Laurence Aitchison
Variational inference in Bayesian neural networks is usually performed using stochastic sampling which gives very high-variance gradients, and hence slow learning.
no code implementations • 20 Nov 2020 • Ali Unlu, Laurence Aitchison
We develop variational Laplace for Bayesian neural networks (BNNs) which exploits a local approximation of the curvature of the likelihood to estimate the ELBO without the need for stochastic sampling of the neural-network weights.
no code implementations • 4 Oct 2020 • Laurence Aitchison, Adam X. Yang, Sebastian W. Ober
We show that the deep inverse Wishart process gives superior performance to DGPs and infinite BNNs on standard fully-connected baselines.
no code implementations • 24 Sep 2020 • Dylan Holden-Sim, Gavin Leech, Laurence Aitchison
Recent work has identified a number of formally incompatible operational measures for the unfairness of a machine learning (ML) system.
no code implementations • ICLR 2021 • Laurence Aitchison
To get Bayesian neural networks to perform comparably to standard neural networks it is usually necessary to artificially reduce uncertainty using a "tempered" or "cold" posterior.
no code implementations • 13 Aug 2020 • Stoil Ganev, Laurence Aitchison
We currently do not have an understanding of semi-supervised learning (SSL) objectives such as pseudo-labelling and entropy minimization as log-likelihoods, which precludes the development of e. g. Bayesian SSL.
1 code implementation • 17 May 2020 • Sebastian W. Ober, Laurence Aitchison
We consider the optimal approximate posterior over the top-layer weights in a Bayesian neural network for regression, and show that it exhibits strong dependencies on the lower-layer weights.
no code implementations • ICML 2020 • Laurence Aitchison
Recent work has argued that neural networks can be understood theoretically by taking the number of channels to infinity, at which point the outputs become Gaussian process (GP) distributed.
no code implementations • ICLR 2019 • Laurence Aitchison
We formulate stochastic gradient descent (SGD) as a novel factorised Bayesian filtering problem, in which each parameter is inferred separately, conditioned on the corresopnding backpropagated gradient.
3 code implementations • ICLR 2019 • Adrià Garriga-Alonso, Carl Edward Rasmussen, Laurence Aitchison
For a CNN, the equivalent kernel can be computed exactly and, unlike "deep kernels", has very few parameters: only the hyperparameters of the original CNN.
1 code implementation • NeurIPS 2020 • Laurence Aitchison
We formulate the problem of neural network optimization as Bayesian filtering, where the observations are the backpropagated gradients.
1 code implementation • NeurIPS 2019 • Laurence Aitchison
Multi-sample, importance-weighted variational autoencoders (IWAE) give tighter bounds and more accurate uncertainty estimates than variational autoencoders (VAE) trained with a standard single-sample objective.
no code implementations • ICLR 2019 • Laurence Aitchison, Vincent Adam, Srinivas C. Turaga
Each training step for a variational autoencoder (VAE) requires us to sample from the approximate posterior, so we usually choose simple (e. g. factorised) approximate posteriors in which sampling is an efficient computation that fully exploits GPU parallelism.
no code implementations • NeurIPS 2017 • Laurence Aitchison, Lloyd Russell, Adam M. Packer, Jinyao Yan, Philippe Castonguay, Michael Hausser, Srinivas C. Turaga
Population activity measurement by calcium imaging can be combined with cellular resolution optogenetic activity perturbations to enable the mapping of neural connectivity in vivo.
no code implementations • NeurIPS 2014 • Guillaume Hennequin, Laurence Aitchison, Mate Lengyel
Multiple lines of evidence support the notion that the brain performs probabilistic inference in multiple cognitive domains, including perception and decision making.
no code implementations • 4 Oct 2014 • Laurence Aitchison, Jannes Jegminat, Jorge Aurelio Menendez, Jean-Pascal Pfister, Alex Pouget, Peter E. Latham
They then use that uncertainty to adjust their learning rates, with more uncertain weights having higher learning rates.