Results on STL-10 at 500 epoch.

<div><p>We introduce PiCCL (Primary Component Contrastive Learning), a self-supervised contrastive learning framework that utilizes a multiplex Siamese network structure consisting of many identical branches rather than 2 to maximize learning efficiency. PiCCL is simple and light weight,...

وصف كامل

محفوظ في:

التفاصيل البيبلوغرافية
المؤلف الرئيسي:	Yiming Kuang (22120458) (author)
مؤلفون آخرون:	Jianwu Guan (22120461) (author), Hongyun Liu (185898) (author), Fei Chen (13690) (author), Zihua Wang (290751) (author), Weidong Wang (321040) (author)
منشور في:	2025
الموضوعات:	Science Policy Space Science Biological Sciences not elsewhere classified use asymmetric networks maximize learning efficiency intricate pretext tasks image numerous times image augmentation paradigm could hinder performance 3 datasets respectively primary component loss compute loss functions piccl &# 8217 network loss xlink "> various state unique structure take advantage supervised algorithms piccl excels multimodal data light weight introduce piccl computationally lightweight calculated using benchmarked piccl batch size
الوسوم:	إضافة وسم لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!

Results on STL-10 at 500 epoch.

مواد مشابهة