14.00 - 14.45 Alessandro Milazzo, Uppsala University
Dynamic programming principle for classical and singular stochastic control with discretionary stopping
Abstract: The dynamic programming principle (DPP) forms the backbone of the whole stochastic (and deterministic) control theory. Although simple and intuitive in its statement, a rigorous mathematical proof can be rather technical and varies across different classes of stochastic control problems. After an extended introduction on the DPP and the classes of classical and singular stochastic control problems, I will show how the concept of regular conditional probabilities allows us to build a pseudo-Markovian structure which leads to the proof of the DPP for control problems combining classical and singular controls with discretionary stopping.
This talk is based on a joint paper with Tiziano De Angelis (University of Torino).
15.00 - 15.30 Coffee
15.30 - 16.15 Jesper Singh, Umeå University
Estimates of p-harmonic functions in planar sectors
Abstract: We are solving the p-Laplace equation in circular sectors (subsets of R^2) and prove a result for the p-harmonic measure that enables us to prove a Phragmen-Lindelöf type of theorem in these sectors.
16.15 - 17.00 Fredrik Ohlsson, Umeå University
Equivariance versus Augmentation for Learning on the Sphere
Abstract: We analyze the role of rotational equivariance in convolutional neural networks (CNNs) applied to spherical data. We compare the performance of baseline architectures of group equivariant networks and standard non-equivariant CNNs trained with an increasing amount of data augmentation. For the task of image.
Speakers: Pekka Nieminen, University of Turku, Alessandro Milazzo, Uppsala University, Jesper SIngh, Umeå University and Fredrik Ohlsson, Umeå University
