Explore model calibration techniques in this 25-minute conference talk from DSC Adria 2024 in Zagreb, where Tin Ferkovic delves into methods for aligning confidence predictions with output probabilities. Learn about both simple and advanced calibration approaches, from basic classification problems in text classification and named entity recognition to sophisticated techniques like TrOCR for generative tasks. Discover how to effectively balance discriminative and generative scenarios to improve model performance and reliability.
Model Calibration Techniques for Discriminative and Generative Tasks
Data Science Conference via YouTube
Overview
Syllabus
Model Calibration | Tin Ferkovic | DSC ADRIA 24
Taught by
Data Science Conference
Related Courses
-
Build custom text analytics solutions
-
Develop natural language processing solutions with Azure AI Services
5.0 -
Generative AI - Opportunities and Security Challenges
-
Generative and Discriminative Models in Machine Learning - Lecture 27B
-
Verif.ai - A Trustworthy Scientific Generative Search Engine
-
Understanding TIS.AI's Communication and Language Model Applications
