In this talk, we shall propose a relaxed control regularization with general exploration rewards to design robust feedback controls for multi-dimensional continuous-time stochastic exit time problems. We establish that the regularized control problem admits a H\”{o}lder continuous feedback control, and demonstrate that both the value function and the feedback control of the regularized control problem are Lipschitz stable with respect to parameter perturbations. Moreover, we show that a pre-computed feedback relaxed control has a robust performance in a perturbed system, and derive a first-order sensitivity equation for both the value function and optimal feedback relaxed control. These stability results provide a theoretical justification for recent reinforcement learning heuristics that including an exploration reward in the optimization objective leads to more robust decision making. We finally prove first-order monotone convergence of the value functions for relaxed control problems with vanishing exploration parameters, which subsequently enables us to construct the pure exploitation strategy of the original control problem based on the feedback relaxed controls. This is joint work with Christoph Reisinger (available at https://arxiv.org/abs/2001.03148).
 

We adopt Deep Reinforcement Learning algorithms to design trading strategies for continuous futures contracts. Both discrete and continuous action spaces are considered and volatility scaling is incorporated to create reward functions which scale trade positions based on market volatility. We test our algorithms on the 50 most liquid futures contracts from 2011 to 2019, and investigate how performance varies across different asset classes including commodities, equity indices, fixed income and FX markets. We compare our algorithms against classical time series momentum strategies, and show that our method outperforms such baseline models, delivering positive profits despite heavy transaction costs. The experiments show that the proposed algorithms can follow large market trends without changing positions and can also scale down, or hold, through consolidation periods.
The full-length text is available at https://arxiv.org/abs/1911.10107.
 

In this talk we will discuss a data-driven approach based on neural networks (NN) for calibrating financial asset price models. Determining optimal values of the model parameters is formulated as training hidden neurons within a machine learning framework, based on available financial option prices. The framework consists of two parts: a forward pass in which we train the weights of the NN off-line, valuing options under many different asset model parameter settings; and a backward pass, in which we evaluate the trained NN-solver on-line, aiming to find the weights of the neurons in the input layer. We will show how the same data-driven approach can be used to estimate the Black-Scholes implied volatility and dividend yield for American options in a fast and robust way. We then discuss the complexity of the optimization problem through an analysis of the loss surface of the neural network. We finally will present some numerical examples which show that neural networks can be an efficient and reliable technique for the calibration of financial assets and the extraction of implied information.

The notion of emergence is at the core of many of the most challenging open scientific questions, being so much a cause of wonder as a perennial source of philosophical headaches. Two classes of emergent phenomena are usually distinguished: strong emergence, which corresponds to supervenient properties with irreducible causal power; and weak emergence, which are properties generated by the lower levels in such "complicated" ways that they can only be derived by exhaustive simulation. While weak emergence is generally accepted, a large portion of the scientific community considers causal emergence to be either impossible, logically inconsistent, or scientifically irrelevant.

In this talk we present a novel, quantitative framework that assesses emergence by studying the high-order interactions of the system's dynamics. By leveraging the Integrated Information Decomposition (ΦID) framework [1], our approach distinguishes two types of emergent phenomena: downward causation, where macroscopic variables determine the future of microscopic degrees of freedom; and causal decoupling, where macroscopic variables influence other macroscopic variables without affecting their corresponding microscopic constituents. Our framework also provides practical tools that are applicable on a range of scenarios of practical interest, enabling to test -- and possibly reject -- hypotheses about emergence in a data-driven fashion. We illustrate our findings by discussing minimal examples of emergent behaviour, and present a few case studies of systems with emergent dynamics, including Conway’s Game of Life, neural population coding, and flocking models.
[1] Mediano, Pedro AM, Fernando Rosas, Robin L. Carhart-Harris, Anil K. Seth, and Adam B. Barrett. "Beyond integrated information: A taxonomy of information dynamics phenomena." arXiv preprint arXiv:1909.02297 (2019).
 

Topology optimisation finds the optimal material distribution of a fluid or solid in a domain, subject to PDE and volume constraints. There are many formulations and we opt for the density approach which results in a PDE, volume and inequality constrained, non-convex, infinite-dimensional optimisation problem without a priori knowledge of a good initial guess. Such problems can exhibit many local minima or even no minima. In practice, heuristics are used to obtain the global minimum, but these can fail even in the simplest of cases. In this talk, we will present an algorithm that solves such problems and systematically discovers as many of these local minima as possible along the way.

A big problem in Riemannian geometry is the search for a "best possible" Riemannian metric on a given compact smooth manifold. When the manifold is complex, one very nice metric we could look for is a Kahler-Einstein metric. For compact Kahler manifolds with non-positive first chern class, these were proven to always exist by Aubin and Yau in the 70's. However, the case of positive first chern class is much more delicate, and there are non-trivial obstructions to existence. It wasn't until this decade that a complete abstract characterisation of Kahler-Einstein metrics became available, in the form of K-stability. This is a purely algebro-geometric stability condition, whose equivalence to the existence of a Kahler-Einstein metric in the Fano case is analogous to the Hitchin-Kobayashi correspondence for vector bundles. In this talk, I will cover the definition of K-stability, its relation to Kahler-Einstein metrics, and (time permitting) give some examples of how K-stability is verified or disproved in practice.

 This is a longstanding problem. We will present candidate axioms that we believe form a complete axiom system satisfied by $F_p^{alg}((t))$.  A currently "unfinished" proof will be discussed.   This work is joint with Françoise Delon (Paris) and Arno Fehm (Dresden).