Adam Olsson och Gabriel Lindeby, MPALG
Titel på masterarbetet: Distributed Training for Deep Reinforcement Learning Decoders on the Toric Code
Följ presentationen online
Meeting ID: 647 6075 6824
Handledare och examinator: Mats Granath
The inherently fragile nature of quantum bits, called qubits, cause unavoidable errors in quantum computations. To avoid these errors, qubits need to constantly be monitored and corrected. However, the correction process, called decoder, can not directly observe the error. Instead, the decoder needs to incorporate statistics into the correction process. Recently, deep reinforcement learning has been suggested as a tool for quantum error correction because of its ability to learn advanced control policies from sensory input. These reinforcement learning algorithms require large amounts of data which takes a long time to produce. In this thesis, we implement a distributed framework to train a decoder based on deep reinforcement learning and that produces the same results in less than half of the wall clock time.
Opponenter: Tobias Sandström och Lars Jansson
Online via Zoom