Tensor Networks for Classical and Quantum Machine Learning Tasks

Explore tensor networks for classical and quantum machine learning tasks in this 52-minute conference talk by Dario Poletti from PCS Institute for Basic Science. Delve into the application of matrix product states and operators, the preferred method for studying one-dimensional strongly interacting many-body quantum systems. Discover how this approach allows for the exploration of the most relevant and numerically manageable portion of an exponentially large space while accurately describing correlations between distant parts of a system—a crucial element in machine learning tasks. Learn about a novel machine learning model that utilizes trained matrix product operators for sequence-to-sequence prediction, enabling the forecast of subsequent sequences based on current time step data. Examine the model's applications in both classical problems, such as cellular automata evolution, and quantum challenges, including predicting quantum state evolution under the influence of unknown external environments, potentially non-Markovian in nature.

Dario Poletti: Tensor networks for classical and quantum machine learning tasks