Tensor networks provide a powerful framework for efficiently representing high-dimensional data and many-body quantum states. Endowing tensor networks with a Riemannian manifold structure provides a natural setting for numerical optimization and analysis. A central feature of tensor networks is their gauge freedom, whose characterisation (captured by so-called fundamental theorems) underlies both their intrinsic structure and the design of numerical algorithms. In this work, we study the interaction between the Riemannian manifold structure and the gauge freedom for several families of tensor networks. Using group actions and Riemannian submersions, we establish a Riemannian fundamental theorem for the tensor network families studied.
This chapter begins with a few remarks indicating briefly the motivation for studying quantum theory from a geometrical point of view, and in particular for the development of a non-linear geometrical quantum theory. It considers the case of an elementary quantum mechanical system consisting of a spin one-half particle in a constant magnetic field. The successes of quantum mechanics have been so wide and varied, that one must be wary of making any gross alteration to the structure of the theory. The connection with twistor theory goes perhaps one level deeper, since the particular geometrical characteristics of quantum theory that emerge as being relevant have to do precisely with the complex analytic structure of the quantum mechanical state manifold. A Kahler manifold can either be regarded as a real Riemannian manifold that is endowed also with a complex structure, or as a complex manifold that is also endowed with a Riemannian structure.