Search Talks

Causal reasoning is vital for effective reasoning in science and medicine. In medical diagnosis, for example, a doctor aims to explain a patient’s symptoms by determining the diseases causing them. This is because causal relations---unlike correlations---allow one to reason about the consequences of possible treatments. However, all previous approaches to machine-learning assisted diagnosis, including deep learning and model-based Bayesian approaches, learn by association and do not distinguish correlation from causation. I will show that these approaches systematically lead to incorrect diagnoses. I will outline a new diagnostic algorithm, based on counterfactual inference, which captures the causal aspect of diagnosis overlooked by previous approaches and overcomes these issues. I will additionally describe recent algorithms from my group which can discover causal relations from uncontrolled observational data and show how these can be applied to facilitate effective reasoning in medical settings such as deciding how to treat certain diseases.

We study the defect operator product expansion (OPE) of displacement operators in free and interacting conformal field theories using replica methods. We show that as n approaches 1 a contact term can emerge when the OPE contains defect operators of twist d−2. For interacting theories and general states we give evidence that the only possibility is from the defect operator that becomes the stress tensor in the n→1 limit. This implies that the quantum null energy condition (QNEC) is always saturated for CFTs with a twist gap. As a check, we show independently that in a large class of near vacuum states, the second variation of the entanglement entropy is given by a simple correlation function of averaged null energy operators as studied by Hofman and Maldacena. This suggests that sub-leading terms in the the defect OPE are controlled by a defect version of the spin-3 non-local light ray operator and we speculate about the possible origin of such a defect operator. For free theories this contribution condenses to a contact term that leads to violations of QNEC saturation.

In this talk I will discuss some recent results on boundary degrees of freedom (or edge modes), and their description via an extended phase space structure containing extra boundary fields. Motivated by a slight modification of the covariant phase space formalism, I will show how the use of a boundary Lagrangian enables to include the edge modes in the phase space and to obtain their boundary dynamics. This will be illustrated on the example of Maxwell theory, where in addition the edge modes can be understood as contributing to entanglement entropy. The possible extension of this result to gravity will then be mentioned.

In this talk we describe a systematic way of producing stochastic solutions to the Yang-Baxter equation from a “base solution” that is not necessarily stochastic. This gives rise to a wide class of new families of stochastic (Markovian) vertex models with integrable weights. In various cases, these models are more amenable to algebraic analysis than are their non-stochastic counterparts, sometimes leading to new (and at times unexpected) refined asymptotic results for them. In other cases, these stochastic systems are amenable to probabilistic analysis, yielding results that seem less transparent to access algebraically and that can occasionally shed new light on the original (non-stochastic) model. This is partially based on joint work with Alexei Borodin and Alexey Bufetov.