notesum.ai

Published at October 23

Congestion Forecast for Trains with Railroad-Graph-based Semi-Supervised Learning using Sparse Passenger Reports

cs.LG

cs.AI

cs.CV

Released Date: October 23, 2024

Authors: Soto Anno¹, Kota Tsubouchi², Masamichi Shimosaka¹

Aff.: ¹Tokyo Institute of Technology, Tokyo, Japan; ²LY Corporation, Tokyo, Japan

Arxiv: https://arxiv.org/abs/2410.17510v1

Refer to caption

			Ratio of labeled data (%)
Model	Protocol	Graph	$10\%$	$25\%$	$50\%$	$75\%$	$100\%$
Random	-	-	$24.60\pm 0.99$	$25.05\pm 0.42$	$25.81\pm 0.48$	$24.34\pm 0.97$	$25.46\pm 1.02$
MODE	stats.	-	$28.03\pm 1.29$	$31.66\pm 0.93$	$37.34\pm 0.80$	$41.42\pm 0.62$	$44.61\pm 0.51$
SNN	SL	-	$54.94\pm 0.91$	$56.15\pm 1.01$	$56.23\pm 0.68$	$57.42\pm 1.80$	$58.75\pm 0.63$
LP (zhu:icml2003, )	SSL	natural	$51.99\pm 1.37$	$52.92\pm 1.19$	$54.42\pm 1.33$	$56.01\pm 1.44$	$56.84\pm 1.34$
LS (zhou:nips03, )	SSL	natural	$51.99\pm 1.37$	$52.92\pm 1.19$	$54.42\pm 1.33$	$56.02\pm 1.45$	$56.83\pm 1.34$
LP-DSSL (iscen:cvpr2019, ) (SOTA)	SSL	descriptor	$49.38\pm 1.36$	$52.97\pm 1.07$	$55.35\pm 0.98$	$57.10\pm 1.52$	$58.88\pm 1.08$
SURCONFORT	SSL	rail	${\bf 56.76\pm 1.93}$	${\bf 58.08\pm 0.95}$	${\bf 59.41\pm 1.37}$	${\bf 60.52\pm 1.43}$	${\bf 60.35\pm 1.34}$