Deep Compression of Sum-Product Networks on Tensor Networks
This work addresses efficiency in probabilistic inference for machine learning, though it appears incremental as it builds on existing SPN and tensor network methods.
The paper tackled the problem of compressing sum-product networks (SPNs) by mapping them onto tensor networks, achieving significant parameter compression with minimal accuracy loss.
Sum-product networks (SPNs) represent an emerging class of neural networks with clear probabilistic semantics and superior inference speed over graphical models. This work reveals a strikingly intimate connection between SPNs and tensor networks, thus leading to a highly efficient representation that we call tensor SPNs (tSPNs). For the first time, through mapping an SPN onto a tSPN and employing novel optimization techniques, we demonstrate remarkable parameter compression with negligible loss in accuracy.