KerPrint: Local-Global Knowledge Graph Enhanced Diagnosis Prediction for Retrospective and Prospective Interpretations

نویسندگان

چکیده

While recent developments of deep learning models have led to record-breaking achievements in many areas, the lack sufficient interpretation remains a problem for specific applications, such as diagnosis prediction task healthcare. The previous knowledge graph(KG) enhanced approaches mainly focus on clinically meaningful representations, importance medical concepts, and even paths from inputs labels. However, it is infeasible interpret prediction, which needs consider different various relationships, time-effectiveness triples patient contexts. More importantly, retrospective prospective interpretations disease processes are valuable clinicians patients' confounding diseases. We propose KerPrint, novel KG approach tackle these problems. Specifically, we time-aware attention method solve decay over time trustworthy interpretation. also element-wise select candidate global using comprehensive representations local validate effectiveness our KerPrint through an extensive experimental study real-world dataset public dataset. results show that proposed not only achieves significant improvement knowledge-enhanced methods but gives interpretability both views.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Global and Local Graph Modifiers

We define two modal logics that allow to reason about modifications of graphs. Both have a universal modal operator. The first one only involves global modifications (of some state label, or of some edge label) everywhere in the graph. The second one also allows for modifications that are local to states. The global version generalizes logics of public announcements and public assignments, as w...

متن کامل

Text-Enhanced Representation Learning for Knowledge Graph

Learning the representations of a knowledge graph has attracted significant research interest in the field of intelligent Web. By regarding each relation as one translation from head entity to tail entity, translation-based methods including TransE, TransH and TransR are simple, effective and achieving the state-of-the-art performance. However, they still suffer the following issues: (i) low pe...

متن کامل

Formulas, and Graph-theoretic Interpretations

We derive a closed-form expression for all genus 1 Hurwitz numbers, and give a simple new graph-theoretic interpretation of Hurwitz numbers in genus 0 and 1. (Hurwitz numbers essentially count irreducible genus g covers of the sphere, with arbitrary specified branching over one point, simple branching over other specified points, and no other branching. The problem is equivalent to counting tra...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: Proceedings of the ... AAAI Conference on Artificial Intelligence

سال: 2023

ISSN: ['2159-5399', '2374-3468']

DOI: https://doi.org/10.1609/aaai.v37i4.25667