Uncertainty and Feature Learning with Kernel Machines for ATR

Abstract

In this grant, we apply a recently developed uncertainty principle inspired by information theoretic learning to applications of interest to the Navy. Our uncertainty measure is non-intrusive to the training process of point-wise prediction models, and has the great advantage of being much faster to compute than traditional techniques. Hence, it can be potentially used for online applicationsin computational restricted environments. We propose to improve the man-machine interface for machine learning operators using uncertainty with the goal of improving the acceptability and trust of ML to effectively boost performance. Another important application is to carefully select exemplars to be added to ATR training sets to improve the performance of ML algorithms to save time and money when creating the databases for the Navy missions.Furthermore, we will also apply a recently developed close form solution to nonlinear prediction problems that will avoid the need for the huge amounts of data required to train conventional machine learning models. We called this novel algorithm the Functional Wiener Filter (FWF). In testing mode, the complexity is O(K2) where K is the order of the filter, and can be easily implemented in ultra-low power hardware. With these properties, we can automatically configure classifiers to any specific scenario. This is a major achievement andwe plan to extend it for images such that can quantify sea floor texture (spatial statistics) that can be used as contextual information for ATR to avoid many of the false alarms. Better understanding of the RKHS for classification, preserving explainability, and reducing computation w.r.t. deep learning will also be crucial for ATR and many other applications ofautonomy.Approved for Public Release.

Document Details

Document Type
DoD Grant Award
Publication Date
Apr 10, 2025
Source ID
N000142512223

Entities

People

  • José Príncipe

Organizations

  • Office of Naval Research
  • United States Navy
  • University of Florida

Tags

Fields of Study

  • Computer science

Readers

  • Calculus or Mathematical Analysis
  • Distributed Systems and Data Platform Development
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - Machine Learning Algorithms
  • AI & ML - Neural Networks