BEGIN:VCALENDAR PRODID:-//eluceo/ical//2.0/EN VERSION:2.0 CALSCALE:GREGORIAN BEGIN:VEVENT UID:520bff32250f8c016fe6efac9cf3852d DTSTAMP:20260411T214010Z SUMMARY:New Trends in Machine Learning for Finance DESCRIPTION:The event intends to bring together leading researchers from ac ademia and financial industry at the intersection of machine learning and finance. The workshop centres around the application and development of no vel machine learning methods for hedging\, market microstructure and marke t scenario generation. \nSpeakers\n\nAmira Akkari (J.P. Morgan)\nRomuald E lie (Deepmind & Université Gustave Eiffel)\nBlanka Horvath (University of Oxford)\nAntoine Jacquier (Imperial College London)\nSemyon Malamud (Éco le polytechnique fédérale de Lausanne)\nHao Ni (University College Londo n)\nJohannes Ruf (London School of Economics)\nJosef Teichmann (ETH Züric h)\nNicholas Westray (Alliance Bernstein and New York University)\n\n \nT he workshop is supported by the Cecilia Tanner Research Funding Scheme.\n\ n \n\n\n\n\n09:10-09:15\nOpening\n\n\n09:15-09:55\nAntoine Jacquier: Rand om neural networks for rough volatility\n\n\n09:55-10:35\nJohannes Ruf: He dging with linear regressions and neural networks\n\n\n10:35-11:00\nCoffee break\n\n\n11:00-11:40\nJosef Teichmann: (Learning) Strategies for ergodi c robust optimal asymptotic growth under stochastic volatility\n\n\n11:40- 12:20\nRomuald Elie: Learning equilibria in mean field games\n\n\n12:20-13 :30\nLunch break\n\n\n13:30-14:10\n\nNicholas Westray: The Informational C ontent of Cross-Sectional Multi-level Order Flow Imbalance in US Equity Ma rkets\n\n\n\n14:10-14:50\nAmira Akkari: Deep Hedging & Deep Bellman Hedgin g\n\n\n14:50-15:15\nCoffee break\n\n\n15:15-15:55\nHao Ni: PCF-GAN: gener ating sequential data via the characteristic function of measures on the p ath space\n\n\n15:55-16:35\nSemyon Malamud: Complexity in Factor Pricing M odels\n\n\n16:35-16:45\nBreak\n\n\n16:45-17:25\nBlanka Horvath: Pathwise m ethods and generative models for pricing and trading\n\n\n\n\n \n\nTitles and Abstracts\n\nAntoine Jacquier (Imperial College London)\nTitle: Rando m neural networks for rough volatility\nAbstract: The classical Feynman-Ka c bridge between Markovian SDEs and PDEs has recently been extended in the context of rough (i.e. non-Markovian) stochastic volatility models\, givi ng rise to path-dependent PDEs. The latter\, however\, lack the numerical analysis foundations their finite-dimensional counterparts have. A naïve discretisation forces one to deal with a high-dimensional PDE\, notoriousl y hard to solve numerically. We focus here on recent developments by Hure- Pham-Warin and Bayer-Qiu-Yao\, where the classical backward resolution tec hnique is combined with neural networks to estimate both the solution and its gradient. Not only does this approach successfully diminish the curse of dimensionality\, but is also shown to be more effective in both accurac y and computational efficiency than existing Euler-based approaches. Our c ontribution is to replace their neural networks by reservoir networks (fol lowing the steps developed by Gonon)\, leading to simple least-square mini misation problems instead of large ML-type training considerations. This i s a joint work with Zan Zuric (Imperial College London).\n\nJohannes Ruf  (London School of Economics)\nTitle: Hedging with linear regressions and neural networks\nAbstract: We study the use of neural networks as nonparam etric estimation tools for the hedging of options. To this end\, we design a network\, named HedgeNet\, that directly outputs a hedging strategy giv en relevant features as input. This network is trained to minimise the hed ging error instead of the pricing error. Applied to end-of-day and tick pr ices of S&P 500 and Euro Stoxx 50 options\, the network is able to reduce the mean squared hedging error of the Black-Scholes benchmark significantl y. We illustrate\, however\, that a similar benefit arises by a simple lin ear regression model that incorporates the leverage effect. Joint work wit h Weiguan Wang.\n\nJosef Teichmann (ETH Zürich)\nTitle: (Learning) Strate gies for ergodic robust optimal asymptotic growth under stochastic volatil ity\nAbstract: We consider an asymptotic robust growth problem under model uncertainty and in the presence of (non-Markovian) stochastic covariance. We fix two inputs representing the instantaneous covariance for the asset process $X$\, which depends on an additional stochastic factor process $Y $\, as well as the invariant density of $X$ together with $Y$. The stochas tic factor process $Y$ has continuous trajectories but is not even require d to be a semimartingale. Our setup allows for drift uncertainty in $X$ an d model uncertainty for the local dynamics of $Y$. This work builds upon a recent paper of Kardaras & Robertson\, where the authors consider an anal ogous problem\, however\, without the additional stochastic factor process . Under suitable\, quite weak assumptions we are able to characterize the robust optimal trading strategy and the robust optimal growth rate. The op timal strategy is shown to be functionally generated and\, remarkably\, do es not depend on the factor process $Y$. Our result provides a comprehensi ve answer to a question proposed by Fernholz in 2002. Mathematically\, we use a combination of partial differential equation (PDE)\, calculus of var iations and generalized Dirichlet form techniques. We also point towards M L ways to illustrate the result. Joint work with David Itkin\, Benedikt Ko ch\, Martin Larsson.\nThe theoretical results are accompanied by generativ e adversarial learning approaches for robust strategies (joint work with F lorian Krach and Hanna Wutte).\n\n\nRomuald Elie (Deepmind & Université G ustave Eiffel)\nTitle: Learning equilibria in mean field games\nAbstract: We will present different approaches and algorithms for learning equilibri a in mean field games. In particular\, we will consider frameworks where u niqueness of Nash does not hold\, and see how one can approximate alternat ive solution concepts\, such as Correlated or Coarse correlated equilibria . Applications such as animal flocking\, vehicle routing as well as connec tions with exploration problems in Reinforcement learning will also be dis cussed.\n\nNicholas Westray (Alliance Bernstein and New York University)\n Title: The Informational Content of Cross-Sectional Multi-level Order Flow Imbalance in US Equity Markets\nAbstract: In this talk we discuss the imp ortance of different types of Order Flow Imbalance (OFI) for contemporaneo us return prediction in the US Equity market. We consider multilevel OFI\, built from the deeper layers of the order book as well as cross sectional OFI built from the imbalances of other stocks. In the multi-level OFI cas e we provide a Bayesian formulation to help identify the best number of le vels to be used in prediction. In the cross sectional OFI case we provide a highly efficient implementation of the well known Automatic Relevance De termination (ARD) method to help identify the number of cross sectional st ocks contributing to the return forecast. We provide practical comments on how to obtain the best model in the cross sectional case\, using the Shap ley value to assess the contribution of various terms to performance. \n\ nAmira Akkari (J.P. Morgan)\nTitle: Deep Hedging & Deep Bellman Hedging\nA bstract: Traditional risk management is based on the Greeks provided by c lassical valuation models. These models typically have simplified dynamics and assume perfect hedge-ability.  As a result\, decisions on when and h ow to hedge are based on traders’ intuition\, experience\, and view on m arket dynamics. With Deep Hedging (DH)\, we go beyond Greek-based hedging and take a new approach to exotics risk management.\nDH formulates the hed ging problem as a reinforcement learning problem and shifts towards the ma chine learning paradigm.  We will present two formulations of DH. In the first formulation\, we solve for the optimal hedging in incomplete markets using a periodic policy search. The model-based policy search approximate s the hedging actions (the policy) using deep neural networks. In the seco nd formulation\, we solve the more rigorous dynamic programming problem un der a Deep Bellman formulation\, where the deep hedging problem is express ed into the infinite time horizon through a recursive Bellman representati on.  This can then be solved numerically by adapting techniques from deep reinforcement learning to give a risk-averse actor critic algorithm. The actor gives the optimal hedge and the critic gives the utility-indifferenc e price of the portfolio.\n\nHao Ni (University College London)\nTitle: PC F-GAN: generating sequential data via the characteristic function of measu res on the path space\nAbstract: Implicit Generative Models (IGMs) have de monstrated a superior capacity in generating high-fidelity samples from th e high-dimensional space\, especially for static image data. However\, the se methods struggle to capture the temporal dependence of joint probabilit y distributions induced by time-series data. To tackle this issue\, we dir ectly compare the path distributions via the characteristic function of me asures on the path space (PCF) from rough path theory\, which uniquely cha racterises the law of stochastic processes. The distance metric via PCF en joyed several theoretical properties\, and it also is linked with the MMD loss on the path space. Furthermore\, the PCF loss can be optimised based on the path distribution by learning the optimal unitary representation of PCF\, which avoids the need for manual kernel selection and improves test power. We validate the effectiveness of the proposed PCF-GAN on several b enchmarking datasets\, such as rough volatility data and empirical financi al data.\n\nSemyon Malamud (École polytechnique fédérale de Lausanne)\n Title: Complexity in Factor Pricing Models\nAbstract: We theoretically cha racterize the behavior of machine learning asset pricing models.  We prov e that expected out-of-sample model performance – in terms of SDF Sharpe ratio and average pricing errors – is improving in model parameterizati on (or “complexity”).  Our results predict that the best asset pricin g models (in terms of expected out-of-sample performance) have an extremel y large number of factors (more than the number of training observations o r base assets).  Our empirical findings verify the theoretically predicte d “virtue of complexity” in the cross-section of stock returns and fin d that the best model combines tens of thousands of factors. We also deriv e the feasible Hansen-Jagannathan (HJ) bound: The maximal Sharpe ratio ach ievable by a feasible portfolio strategy. The infeasible HJ bound massivel y overstates the achievable maximal Sharpe ratio due to a complexity wedge that we characterize. Joint work with Antoine Didisheim\, Shikun Ke and B ryan Kelly.\n\n\nBlanka Horvath (University of Oxford)\nTitle: Pathwise me thods and generative models for pricing and trading\nAbstract: The deep h edging framework as well as related deep trading setups have opened new ho rizons for solving hedging problems under a large variety of models and ma rket conditions. In this setting\, generative models and pathwise methods rooted in rough paths have proven to be a powerful tool from several persp ectives. At the same time\, any model – a traditional stochastic model o r a market generator – is at best an approximation of market reality\, p rone to model-misspecification and estimation errors. In a data-driven set ting\, especially if sample sizes are limited by constraints\, the latter issue becomes even more prevalent\, which we demonstrate in examples. This raises the question\, how to furnish a modelling setup (for deriving a st rategy) with tools that can address the risk of the discrepancy between mo del and market reality\, ideally in a way that is automatically built in t he setting. A combination of classical and new tools yields insights into this matter.\n URL:https://www.imperial.ac.uk/events/159297/new-trends-in-machine-learning -for-finance/ DTSTART;TZID=Europe/London:20230330T090000 DTEND;TZID=Europe/London:20230330T180000 LOCATION:The Ballroom\, 58 Prince's Gate\, South Kensington Campus\, Imperi al College London\, London\, SW7 2PR\, United Kingdom END:VEVENT BEGIN:VTIMEZONE TZID:Europe/London BEGIN:DAYLIGHT DTSTART:20230330T090000 TZNAME:BST TZOFFSETTO:+0100 TZOFFSETFROM:+0100 END:DAYLIGHT END:VTIMEZONE END:VCALENDAR