QuantAlps
Registration is now closed for the whole conference.
https://app.livestorm.co/y-spot/interplay-between-neural-and-tensor-networks?type=detailed
The first day will be devoted to tutorial sessions opened to a larger public to put everyone on the same page. There will be half a day devoted to an introduction to tensor networks by Miles Stoudenmire and half a day devoted to an introduction to modern neural networks by Lei Wang.
The rest of the programme is still in construction.
Two very different mathematical objects are getting increasingly popular to represent objects in a compressed way: neural networks and tensor networks. Both are mathematical functions that can approximate various things e.g. the probability distribution of text (in a large language model), the solution of a partial differential equation or the ground state of a quantum many-body system. Both use some form of lossy compression, albeit of very different nature, and in both cases, it is not necessarily easy to know which object will compress well and which won’t. There are in fact intimate connections between the two types of networks and both rely on heavily optimized linear algebra libraries. There are on the other hand important differences as well: what one lacks in expressivity (pretty much everything can be fitted with a neural network; tensor networks are perhaps less flexible), may be complemented by advanced algorithms (tensor networks provide efficient routes to calculate many things without the need for stochastic methods).
This workshop will gather experts of both fields with a focus on the application to quantum many-body physics: neural quantum states on one hand and tensor network states on the other. The goal, besides discussing the latest developments and bridging the two communities together, is to explore approaches that somehow lie at the interplay between the two fields.