Webinar FaIR Deep Learning in Finance: From Implementation to Regulation – Replay Last modification: 10/01/2021 Replay AI-based Asset & Risk Management - 27 September 2021 The ILB FaIR webinar took place on 27 September, in partnership with EDF, ACPR and The Alan Turing Institute. AI computation of trading and hedging strategies has opened up new opportunities, including the ability to solve high-dimensional problems, management of constraints (liquidity, transaction costs, proxy hedging), and a more flexible choice of the criterion to be optimised. In this session, we will present the latest improvements to these methods and their operational use. See video Replay Generative Methods for Simulations and Risk Management - 28 September 2021 Generative methods (GANs, VAEs, etc.) applied to time series simulations allow the model to be updated in a flexible way without having to spend time designing a new stochastic model. However, the direct application of generative adversarial networks (GANs) to time series is not straightforward. We present recent advances in time series generation and discuss the issues they raise. See video Replay Confidence and Regulation of AI-based Algorithms - 29 September 2021 AI brings many improvements to the financial sector: faster and more flexible, AI algorithms tend to provide better forecasts, simulations or internal controls. However, there is a lack of confidence when it comes to the industrial implementation of AI: insufficient unit testing, lack of theoretical guarantees, data sensitivity. We will discuss how to increase the explainability and trust in AI algorithms, both from the regulators’ and the industry’s perspective. See video × × × Slides Day 1 – AI-based Asset & Risk Management Monday 27/09 (9:00-12:30 AM CET) Arnulf Jentzen (University of Münster & The Chinese University of Hong Kong) Title: Convergence analysis for gradient descent optimization methods in the training of ReLU neural networks – get the slides Josef Teichmann (ETH Zurich) Title: Deep Asset Liability Management – get the slides Joseph Mikael (EDF – Senior Research Engineer) Title: A Quick Overview of EDF’s AI Research and Applications in Finance Related Activities – get the slides Day 2 – Generative Methods for Simulations and Risk Management Tuesday 28/09 (9:00 – 12:30 AM CET) Lukasz Szpruch (University of Edinburgh & Alan Turing Institute) Title: Neural SDEs and their Offsprings in Risk Management – get the slides Blanka Horvath (King’s College & Technical University of Munich & Alan Turing Institute) Title: Kernel Methods in Generative Modelling – get the slides Edmond Lezmi (Amundi, Head of Multi-Asset Quantitative Research) Title: Trading Strategy Backtesting with Boltzmann Machines and Generative Adversarial Networks – get the paper Day 3 – Confidence and Regulation of AI-based Algorithms Wednesday 29/09 (9-12:30 AM CET) Jean-Michel Loubes (Université Toulouse Paul Sabatier) Title: What solutions can be Provided Using Mathematical Tools? – get the slides Stéphane Crépey (Université de Paris) Title: Darwinian Model Risk and Reverse Stress Testing – get the slides Olivier Fliche (ACPR – Head of Fintech/Innovation department) Title: Gouvernance of AI Algorithms in the Financial Sector – get the slides Documentation Autorité de Contrôle Prudentiel et de Résolution (ACPR) documentation 1st report AI (2018) 2nd report AI (governance, 2020) AI and Finance Mondays webinars Additional Information Webinar FaIR: Governance of AI in finance… and beyond – Institut Louis Bachelier Robo-Advising: Less AI and more XAI Chair Good in Tech: Institut Mines Telecom – Sciences Po Chair Finance Digitale Université Paris II – Telecom ParisTech
Replay AI-based Asset & Risk Management - 27 September 2021 The ILB FaIR webinar took place on 27 September, in partnership with EDF, ACPR and The Alan Turing Institute. AI computation of trading and hedging strategies has opened up new opportunities, including the ability to solve high-dimensional problems, management of constraints (liquidity, transaction costs, proxy hedging), and a more flexible choice of the criterion to be optimised. In this session, we will present the latest improvements to these methods and their operational use. See video
Replay Generative Methods for Simulations and Risk Management - 28 September 2021 Generative methods (GANs, VAEs, etc.) applied to time series simulations allow the model to be updated in a flexible way without having to spend time designing a new stochastic model. However, the direct application of generative adversarial networks (GANs) to time series is not straightforward. We present recent advances in time series generation and discuss the issues they raise. See video
Replay Confidence and Regulation of AI-based Algorithms - 29 September 2021 AI brings many improvements to the financial sector: faster and more flexible, AI algorithms tend to provide better forecasts, simulations or internal controls. However, there is a lack of confidence when it comes to the industrial implementation of AI: insufficient unit testing, lack of theoretical guarantees, data sensitivity. We will discuss how to increase the explainability and trust in AI algorithms, both from the regulators’ and the industry’s perspective. See video
