|Monday 6 April 2020|
|Events for day: Wednesday 08 January 2020|
| 11:00 - 12:30 Geometry and Topology Weekly Seminar|
Statistical Stability for Singular Hyperbolic Attractors
A common agreement on the definition of chaos is the sensitive dependence on initial conditions. That means independent of how close two initial conditions are, by letting the system to proceed for a while, the new resulting states of the system are significantly different. In other words, a small error at a starting point will cause a huge difference in the outcome of the system. Since measuring a starting point can not be done accurately, the orbit of states is quite unpredictable. But statistically there is a hope to make a prediction by measuring an observable along orbits of the system. Despite of the alteration of the observable along a ...
16:00 - 17:00 Mathematics Colloquium
A Cryptographic Approach to Robust Learning
Devising classification algorithms that are robust to worst-case perturbations has emerged as a challenging problem in theoretical machine learning. In this work, we study whether there is any learning task for which it is possible to design classifiers that are only robust against polynomial-time adversaries; just like how it is done cryptography. Indeed, numerous cryptographic tasks (e.g. encryption of long messages) are only secure against computationally bounded adversaries. We show that computational limitation of attackers can indeed be useful in robust learning by demonstrating a classifier for a learning task in which computational ...