Recognition: unknown
TensorNetwork for Machine Learning
read the original abstract
We demonstrate the use of tensor networks for image classification with the TensorNetwork open source library. We explain in detail the encoding of image data into a matrix product state form, and describe how to contract the network in a way that is parallelizable and well-suited to automatic gradients for optimization. Applying the technique to the MNIST and Fashion-MNIST datasets we find out-of-the-box performance of 98% and 88% accuracy, respectively, using the same tensor network architecture. The TensorNetwork library allows us to seamlessly move from CPU to GPU hardware, and we see a factor of more than 10 improvement in computational speed using a GPU.
This paper has not been read by Pith yet.
Forward citations
Cited by 2 Pith papers
-
Tensor-Augmented Convolutional Neural Networks: Enhancing Expressivity with Generic Tensor Kernels
A two-layer tensor-augmented CNN reaches 93.7% test accuracy on Fashion-MNIST, matching or exceeding deeper models like VGG-16 and GoogLeNet by using tensors for richer expressivity.
-
SMT-AD: a scalable quantum-inspired anomaly detection approach
SMT-AD applies superposition of bond-dimension-1 matrix product operators with multiresolution Fourier embedding to achieve competitive anomaly detection on standard datasets with linear parameter growth.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.