{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/scalable-gaussian-processes-with-billions-of","title":"Scalable Gaussian Processes with Billions of Inducing Inputs via Tensor Train Decomposition","arxiv_id":"1710.07324","date":"2017-10-19","proceeding":null,"authors":["Pavel Izmailov","Alexander Novikov","Dmitry Kropotov"],"abstract":"We propose a method (TT-GP) for approximate inference in Gaussian Process\n(GP) models. We build on previous scalable GP research including stochastic\nvariational inference based on inducing inputs, kernel interpolation, and\nstructure exploiting algebra. The key idea of our method is to use Tensor Train\ndecomposition for variational parameters, which allows us to train GPs with\nbillions of inducing inputs and achieve state-of-the-art results on several\nbenchmarks. Further, our approach allows for training kernels based on deep\nneural networks without any modifications to the underlying GP model. A neural\nnetwork learns a multidimensional embedding for the data, which is used by the\nGP to make the final prediction. We train GP and neural network parameters\nend-to-end without pretraining, through maximization of GP marginal likelihood.\nWe show the efficiency of the proposed approach on several regression and\nclassification benchmark datasets including MNIST, CIFAR-10, and Airline.","url_abs":"http://arxiv.org/abs/1710.07324v2","url_pdf":"http://arxiv.org/pdf/1710.07324v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"scalable-gaussian-processes-with-billions-of","repo_url":"https://github.com/izmailovpavel/TTGP","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"gaussian-processes","task_name":"Gaussian Processes"},{"task_slug":"variational-inference","task_name":"Variational Inference"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1710.07324","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}