Safety and efficacy of the ChAdOx1 nCoV-19 vaccine (AZD1222) against SARS-CoV-2: an interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK.
Voysey, M Clemens, S Madhi, S Weckx, L Folegatti, P Aley, P Angus, B Baillie, V Barnabas, S Bhorat, Q Bibi, S Briner, C Cicconi, P Collins, A Colin-Jones, R Cutland, C Darton, T Dheda, K Duncan, C Emary, K Ewer, K Fairlie, L Faust, S Feng, S Ferreira, D Finn, A Goodman, A Green, C Heath, P Hill, C Hill, H Hirsch, I Hodgson, S Izu, A Jackson, S Jenkin, D Joe, C Kerridge, S Koen, A Kwatra, G Lazarus, R Lawrie, A Lelliott, A Libri, V Lillie, P Mallory, R Mendes, A Milan, E Minassian, A McGregor, A Morrison, H Mujadidi, Y Nana, A O'Reilly, P Padayachee, S Pittella, A Plested, E Pollock, K Ramasamy, M Rhead, S Schwarzbold, A Singh, N Smith, A Song, R Snape, M Sprinz, E Sutherland, R Tarrant, R Thomson, E Török, M Toshner, M Turner, D Vekemans, J Villafana, T Watson, M Williams, C Douglas, A Hill, A Lambe, T Gilbert, S Pollard, A Oxford COVID Vaccine Trial Group Lancet (London, England) volume 397 issue 10269 99-111 (08 Jan 2021) doi:10.1016/s0140-6736(20)32661-1
REACT-1 round 7 interim report: fall in prevalence of swab-positivity in England during national lockdown
Riley, S Eales, O Walters, C Wang, H Ainslie, K Atchison, C Fronterre, C Diggle, P Ashby, D Donnelly, C Cooke, G Barclay, W Ward, H Darzi, A Elliott, P 2020.11.30.20239806 (02 Dec 2020) doi:10.1101/2020.11.30.20239806
Tue, 23 Feb 2021
14:00
Virtual

Dense for the price of sparse: Initialising deep nets with efficient sparse affine transforms

Ilan Price
(Mathematical Institute)
Abstract

That neural networks may be pruned to high sparsities and retain high accuracy is well established. Recent research efforts focus on pruning immediately after initialization so as to allow the computational savings afforded by sparsity to extend to the training process. In this work, we introduce a new `DCT plus Sparse' layer architecture, which maintains information propagation and trainability even with as little as 0.01% trainable kernel parameters remaining. We show that standard training of networks built with these layers, and pruned at initialization, achieves state-of-the-art accuracy for extreme sparsities on a variety of benchmark network architectures and datasets. Moreover, these results are achieved using only simple heuristics to determine the locations of the trainable parameters in the network, and thus without having to initially store or compute with the full, unpruned network, as is required by competing prune-at-initialization algorithms. Switching from standard sparse layers to DCT plus Sparse layers does not increase the storage footprint of a network and incurs only a small additional computational overhead.

--

A link for this talk will be sent to our mailing list a day or two in advance.  If you are not on the list and wish to be sent a link, please contact @email.

Tue, 26 Jan 2021
14:00
Virtual

Preconditioners for computing multiple solutions in three-dimensional fluid topology optimisation

John Papadopoulos
(Mathematical Institute)
Abstract

Topology optimisation finds the optimal material distribution of a fluid or solid in a domain, subject to PDE, volume, and box constraints. The optimisation problem is normally nonconvex and can support multiple local minima. In recent work [1], the authors developed an algorithm for systematically discovering multiple minima of two-dimensional problems through a combination of barrier methods, active-set strategies, and deflation. The bottleneck of the algorithm is solving the Newton systems that arise. In this talk, we will present preconditioning methods for these linear systems as they occur in the topology optimization of Stokes flow. The strategies involve a mix of block preconditioning and specialized multigrid relaxation schemes that reduce the computational work required and allow the application of the algorithm to three-dimensional problems.

[1] “Computing multiple solutions of topology optimization problems”, I. P. A. Papadopoulos, P. E. Farrell, T. M. Surowiec, 2020, https://arxiv.org/abs/2004.11797

 

A link for this talk will be sent to our mailing list a day or two in advance.  If you are not on the list and wish to be sent a link, please contact @email.

The nascent coffee ring: how solute diffusion counters advection
Moore, M Vella, D OLIVER, J Journal of Fluid Mechanics
An exact method for quantifying the reliability of end-of-epidemic declarations in real time
Parag, K Donnelly, C Jha, R Thompson, R PLOS Computational Biology volume 16 issue 11 e1008478 (30 Nov 2020) doi:10.1371/journal.pcbi.1008478
State-level tracking of COVID-19 in the United States
Unwin, H Mishra, S Bradley, V Gandy, A Mellan, T Coupland, H Ish-Horowicz, J Vollmer, M Whittaker, C Filippi, S Xi, X Monod, M Ratmann, O Hutchinson, M Valka, F Zhu, H Hawryluk, I Milton, P Ainslie, K Baguelin, M Boonyasiri, A Brazeau, N Cattarino, L Cucunuba, Z Cuomo-Dannenburg, G Dorigatti, I Eales, O Eaton, J van Elsland, S FitzJohn, R Gaythorpe, K Green, W Hinsley, W Jeffrey, B Knock, E Laydon, D Lees, J Nedjati-Gilani, G Nouvellet, P Okell, L Parag, K Siveroni, I Thompson, H Walker, P Walters, C Watson, O Whittles, L Ghani, A Ferguson, N Riley, S Donnelly, C Bhatt, S Flaxman, S Nature Communications volume 11 issue 1 6189 (03 Dec 2020) doi:10.1038/s41467-020-19652-6
Subscribe to