This paper introduces Artificial Special Intelligence, a new concept enabling error-free training of Machine Learning models for classification, preventing repeated mistakes. The method was successfully applied to 18 MedMNIST biomedical datasets, achieving perfection in most cases despite some double-labeling issues.
This paper introduces Soft Silhouette Loss, a novel differentiable objective for deep learning, inspired by the classical silhouette coefficient. It aims to learn discriminative representations by enforcing intra-class compactness and inter-class separation more efficiently than existing metric learning approaches.
This content explains how to use ROC curves and AUC to fairly compare classification models by assessing performance across all possible thresholds. It details what they are, how to interpret them, and when to use them instead of other metrics, including common misconceptions.
This paper reviews the Receiver Operating Characteristic (ROC) curve, a common metric for evaluating binary classifier performance. It formalizes the probabilistic interpretation of the Area Under the Curve (AUC) and provides bounds for its accuracy.
PLACO is a multi-stage framework designed for cost-effective performance in Human-AI teams, particularly for classification tasks. The paper addresses the crucial combination of human and model output, building upon prior work that used Bayes rule.
The author shares a prompt engineering tip for local LLMs, suggesting to require explicit in-domain terms for "YES" answers to reduce false positives in classification. This technique was applied while building a Reddit signal classifier agent for community intelligence, integrating various LLMs and a Bayes classifier.