Title
An embarrassingly simple approach to neural multiple instance classification
Abstract
•Neural networks trained using ranking-like loss function performs multiple instance learning.•Proposed method shows good generalization performance even for small training sets.•Proposed loss can be used with any architecture of choice without addition of any specialized layers or connections.
Year
DOI
Venue
2019
10.1016/j.patrec.2019.10.022
Pattern Recognition Letters
Keywords
Field
DocType
Machine Learning,Classification,Multiple Instance Learning,Neural Networks
MNIST database,Ranking,Pattern recognition,Convolutional neural network,Image processing,Artificial intelligence,Artificial neural network,Machine learning,Python (programming language),Mathematics
Journal
Volume
ISSN
Citations 
128
0167-8655
0
PageRank 
References 
Authors
0.34
0
2
Name
Order
Citations
PageRank
Amina Asif155.51
Fayyaz ul Amir Afsar Minhas2279.37