At #ICML2023? Don't miss CDT student Nicola Muca Cirone presenting their research carried out with Maud Lemercier, and @CrisSalvi
Neural signature kernels as infinite-width-depth-limits of controlled ResNets
https://t.co/YlL7sOPsrJ
It has fast compile times, fast runtimes, support for general linear operators, efficient implementations of new and existing algorithms in JAX, and more! 2/2
CDT student @packquickly has been working on open-source software with Google that is now available via https://t.co/vmwtdddUI5.
Lineax is a JAX library for fast linear solves and least squares. 1/2
CDT student Debbie Miori will chair a #SIAMFM23 session on Microstructure and Execution (June 8). The session will include CDT student Lorenzo Lucchese The Short-Term Predictability of Mid-Price Returns in Order Book Markets: A #DeepLearning Perspective. https://t.co/KgEmhgzFkU
New from CDT student Bernat Bassols-Cornudella @ImperialMaths
Binomial Cayley Graphs and Applications to Dynamics on Finite Spaces
The paper is co-written with Francesco Viganò of the @EPSRC CDT in Geometry and Number Theory #Combinatorics#PhD
https://t.co/TkwL6dPWZB
This work follows previous results on Model Calibration (or Model Generation...) via Optimal Transport in various settings such as local volatility, local stochastic volatility, path dependent exotic products, VIX SPX calibration, and shows the versatility of this approach. 3/3
CDT Student Benjamin Joseph, Jan Obloj @oxunimaths & Gregoire Loeper @BNPParibas have been working on model calibration via #OptimalTransport.
Calibration of Local Volatility Models with Stochastic Interest Rates using Optimal Transport https://t.co/e286VAxIj3 1/3
They have addressed the case of calibration of equity products with stochastic rates, for which they give an exact calibration procedure. In their applications the resulting model is a local volatility model, where the state variables are the equity spot and the short rate. 2/3
If you weren't lucky enough to see @DesHigham live at our Conference of Mathematics of Random Systems @ICMS_Edinburgh we're happy to share this video of his excellent Public Lecture - #DeepLearning: What Could Go Wrong? - 25/04/23 https://t.co/ACcekDvhBE
Mental Health Awareness Week runs from 15–21 May.
Some events are now fully booked, but there are still plenty of opportunities to get involved. Don't forget to sign-up for daily tips emails.
👀 View the full programme: https://t.co/ycs9jtLFGN
#mentalhealthawarenessweek
Under 30% of STEM postgraduate research students identify as female. Our first Maryam Mirzakhani Scholar will address the M in STEM.
Marta joins us in the autumn. We look forward to many more. Thanks to @xtxmarkets for principal funding. #WomeninMaths
https://t.co/e7JayF82Ks
In the special case where the activation is the identity, we show that the equation reduces to a linear PDE and the limiting kernel agrees with the signature kernel (https://t.co/XLpQRlcv9o). (4/4)
Congratulations to first year CDT student Nicola Muça Cirone @ImperialMaths on their paper with @CristopherSalvi being accepted at #icml2023 (https://t.co/YlL7sOOUCb).
these architectures converge weakly to Gaussian processes indexed on some spaces of paths and with kernels satisfying certain partial differential equations (PDEs) varying according to the choice of activation function. (3/4)
It's only 5 tweets since Oxford Mathematician Alison Etheridge was last elected member of an organisation. Luckily no-one remembers tweets for more than 5 seconds so we can confidently announce Alison has been elected member of the National Academy of Sciences. @theNASciences
Thanks to our keynote speakers Ilya Chevyrev, Ellen Powell, @eyal_neuman & @SamCMaths for sharing their research with us in a selection of excellent talks. Thanks also to @DesHigham for a hugely popular public talk on #DeepLearning (what could go wrong?) https://t.co/AG371sq1A5
Thanks so much to all who took part in our 2023 Conference on Mathematics of Random Systems @ICMS_Edinburgh. We had excellent talks from our amazing students reflecting the broad range of topics in the CDT. The abstracts can be found on our website https://t.co/g2qTbLbmgg