Decision Boundary
A decision boundary in machine learning defines the separation between different classes in a feature space, with the primary objective being to create a boundary that accurately classifies unseen data. Current research focuses on improving decision boundary robustness and stability, particularly addressing challenges like class imbalance, noisy labels, and adversarial attacks, often employing techniques within neural networks (including convolutional and transformer architectures) and leveraging methods such as mixup, contrastive learning, and distributionally robust optimization. Understanding and controlling the properties of decision boundaries is crucial for building reliable and robust machine learning models across diverse applications, from network traffic classification to anomaly detection and open-set learning.
Papers
Dimension-independent learning rates for high-dimensional classification problems
Andres Felipe Lerma-Pineda, Philipp Petersen, Simon Frieder, Thomas Lukasiewicz
Multiplicative Logit Adjustment Approximates Neural-Collapse-Aware Decision Boundary Adjustment
Naoya Hasegawa, Issei Sato
SCOMatch: Alleviating Overtrusting in Open-set Semi-supervised Learning
Zerun Wang, Liuyu Xiang, Lang Huang, Jiafeng Mao, Ling Xiao, Toshihiko Yamasaki
Finding Patterns in Ambiguity: Interpretable Stress Testing in the Decision~Boundary
Inês Gomes, Luís F. Teixeira, Jan N. van Rijn, Carlos Soares, André Restivo, Luís Cunha, Moisés Santos
Sequential sampling without comparison to boundary through model-free reinforcement learning
Jamal Esmaily, Rani Moran, Yasser Roudi, Bahador Bahrami