Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: On the Unreported-Profile-is-Negative Assumption for Predictive Cheminformatics
(Submitted on 3 Apr 2017 (v1), last revised 7 Aug 2017 (this version, v3))
Abstract: In cheminformatics, compound-target binding profiles has been a main source of data for research. For data repositories that only provide positive profiles, a popular assumption is that unreported profiles are all negative. In this paper, we caution audience not to take this assumption for granted, and present empirical evidence of its ineffectiveness from a machine learning perspective. Our examination is based on a setting where binding profiles are used as features to train predictive models; we show (1) prediction performance degrades when the assumption fails and (2) explicit recovery of unreported profiles improves prediction performance. In particular, we propose a framework that jointly recovers profiles and learns predictive model, and show it achieves further performance improvement. The presented study not only suggests applying matrix recovery methods to recover unreported profiles, but also initiates a new missing feature problem which we called Learning with Positive and Unknown Features.
Submission history
From: Chao Lan [view email][v1] Mon, 3 Apr 2017 15:33:10 GMT (525kb)
[v2] Tue, 11 Jul 2017 14:48:57 GMT (126kb)
[v3] Mon, 7 Aug 2017 19:31:49 GMT (0kb,I)
Link back to: arXiv, form interface, contact.