Tanimoto Random Features for Scalable Molecular Machine Learning

Published: 21 Sept 2023, Last Modified: 11 Jan 2024NeurIPS 2023 posterEveryoneRevisionsBibTeX
Keywords: Tanimoto, Kernel, MinMax, Gaussian process, molecule, chemistry, random features
TL;DR: We propose two kinds of random features for the Tanimoto/Jaccard kernel and apply it to molecule data.
Abstract: The Tanimoto coefficient is commonly used to measure the similarity between molecules represented as discrete fingerprints, either as a distance metric or a positive definite kernel. While many kernel methods can be accelerated using random feature approximations, at present there is a lack of such approximations for the Tanimoto kernel. In this paper we propose two kinds of novel random features to allow this kernel to scale to large datasets, and in the process discover a novel extension of the kernel to real-valued vectors. We theoretically characterize these random features, and provide error bounds on the spectral norm of the Gram matrix. Experimentally, we show that these random features are effective at approximating the Tanimoto coefficient of real-world datasets and are useful for molecular property prediction and optimization tasks. Future updates to this work will be available at http://arxiv.org/abs/2306.14809.
Submission Number: 2872
Loading