Tensor network

A tensor network is a graphical representation of a high-dimensional tensor as a contraction of lower-rank tensors arranged on a graph, providing an efficient ansatz for describing quantum many-body states that would otherwise require exponentially many parameters. The framework originated in condensed matter physics — matrix product states for one-dimensional systems, projected entangled pair states for two dimensions — and has become central to quantum information theory, where it provides a natural language for entanglement, error correction, and holography. The key insight is that physical quantum states are not randomly drawn from Hilbert space; they occupy a structured, low-dimensional manifold that tensor networks can parametrize.

The connection to spin networks is structural: both use graphs to encode quantum information, and both rely on the representation theory of groups to label the local degrees of freedom. In tensor networks, the labels are typically the dimensions of vector spaces; in spin networks, they are the spins of SU(2) representations. The mathematical kinship suggests that the tools developed for one — renormalization, coarse-graining, entanglement entropy — may transfer to the other, and that the boundary between condensed matter and quantum gravity may be less sharp than disciplinary conventions assume.

Tensor networks are the proof that high-dimensional quantum systems are not as complex as they appear. The state space of a hundred-qubit system is larger than the number of atoms in the observable universe, yet the physical states that actually occur — the ground states of local Hamiltonians — occupy a tiny, structured corner of that space. The tensor network is the map of that corner, and the fact that such a map exists at all is one of the deepest facts about quantum mechanics.