University of Birmingham > Talks@bham > Theoretical Physics Seminars > Tensor Networks for Machine Learning (Note venue)

Tensor Networks for Machine Learning (Note venue)

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Mike Gunn.

Tensor networks are a technique originating in quantum many-body physics, which underpin powerful and interesting computational techniques. They can be viewed as a way of storing exponentially big quantum wavefunctions using much fewer parameters. Tensor networks are valued for their ability to be optimized very quickly with adaptive algorithms, and their high degree of interpretability.

If one takes a more general view of tensor networks, they are just a kind of function approximator. This means they could be applied to many interesting problems outside of quantum mechanics too. A particularly interesting setting is modeling real-world, classical data—-machine learning—-where challenges similar to quantum mechanics arise, such as computing joint probabilities of many variables and efficiently computing marginal probabilities.

I will discuss frameworks for applying tensor networks, and tensor network algorithms to perform both supervised and unsupervised learning tasks, and discuss recent progress in this area. One exciting & real possibility is theoretically predicting the outcome of algorithms for training tensor network models based on summary properties of the data.

Another interesting idea is to use quantum circuits identical to tensor networks for machine learning on quantum hardware, allowing transfer of model designs and training methods across these platforms. I will conclude by discussing some future directions where tensor networks could have a significant impact.

This talk is part of the Theoretical Physics Seminars series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


Talks@bham, University of Birmingham. Contact Us | Help and Documentation | Privacy and Publicity.
talks@bham is based on from the University of Cambridge.