From 76e336efbd18dd78c6acfb1f1802aa3d009bf882 Mon Sep 17 00:00:00 2001 From: Yue Liu <41297969+YueLiu-coder@users.noreply.github.com> Date: Thu, 15 Dec 2022 23:45:09 +0800 Subject: [PATCH] Update --- README.md | 36 ++++++++++++++++++++++++++++++++++-- assets/HSAN_result.png | Bin 0 -> 306146 bytes assets/HSAN_tsne.png | Bin 0 -> 376432 bytes 3 files changed, 34 insertions(+), 2 deletions(-) create mode 100644 assets/HSAN_result.png create mode 100644 assets/HSAN_tsne.png diff --git a/README.md b/README.md index 8cff392..c6b0163 100644 --- a/README.md +++ b/README.md @@ -28,27 +28,59 @@ An official source code for paper Hard Sample Aware Network for Contrastive Deep ### Overview
-We propose a novel contrastive deep graph clustering method dubbed Hard Sample Aware Network (HSAN) by introducing a comprehensive similarity measure criterion and a general dynamic sample weighing strategy. Concretely, in our algorithm, the similarities between samples are calculated by considering both the attribute embeddings and the structure embeddings, better-revealing sample relationships and assisting hardness measurement. Moreover, under the guidance of the carefully collected high-confidence clustering information, our proposed weight modulating function will first recognize the positive and negative samples and then dynamically up-weight the hard sample pairs while down-weighting the easy ones. In this way, our method can mine not only the hard negative samples but also the hard positive sample, thus improving the discriminative capability of the samples further. Extensive experiments and analyses demonstrate the superiority and effectiveness of our proposed method. +We propose a novel contrastive deep graph clustering method dubbed Hard Sample Aware Network (HSAN) by introducing a comprehensive similarity measure criterion and a general dynamic sample weighing strategy. Concretely, in our algorithm, the similarities between samples are calculated by considering both the attribute embeddings and the structure embeddings, better-revealing sample relationships and assisting hardness measurement. Moreover, under the guidance of the carefully collected high-confidence clustering information, our proposed weight modulating function will first recognize the positive and negative samples and then dynamically up-weight the hard sample pairs while down-weighting the easy ones. In this way, our method can mine not only the hard negative samples but also the hard positive sample.
sCr~ELN+5
z@VFe^V`F2h;(Gt?o)SPoLBaO IXXoweHgzFt$FcnG1(l_4Wp%
zmYdD+75a*{%U?(#9BZd62Nuu7*{)w8VX;|Y`$W9_3@D>d;nPFLGFWyos)c`C$amba
zMViWcW~)Y;j9~r_TWGBxgd0oDn_hl(w)CAAnLK3@zu7;jvwuG=BTTKr)11*e1dg=X
zFJ#n2aM8OIrM7EW4j?Ex_9>FI86f@~b+&9NdF5+Y QC(b$8^-Hzz<(((B##GME};l)
zhc{F2c=o!XbaFa&!g=z*N3j=slRVlCDz@d7p-hP>IT~5rQ?<%y&|^8_cO#3!n~&fY
zhr(FoIJ{-ImSRy+O8*S88;m}#c=^l_#QTI-^ LVQ-p
z4o?MO1JJ84*TXy0lE&w!TZQDHqHTX|feB%{MSw7`NKj(yUUcz5s{GeE(-r^+*AOSYVCngBx2i#Q)LVC
zpszU`#Rc#9XUfoo*B=Mcz+(A!&~SGCFCU&Wd;_vU7SH
z-(5z7WNX@gj>V-XYd