Visualization for Machine Learning

• Peter Xenopoulos New York University
• Luis Gustavo Nonato USP
• Claudio Silva New York University

Resumo

As machine learning has increased in popularity, visualization has taken an important role in analyzing and communicating aspects of machine learning models. Increasingly, visualization techniques are being used across a broad set of domains and in business-critical use cases. Oftentimes, these visualizations convey non-trivial machine learning concepts, utilize complex visual representations, or demand user interaction. In this tutorial, we seek to provide a foundational understanding, to a broad audience, of the ways in which we can use visualization for machine learning tasks. In particular, we detail visual techniques for model assessment, model understanding, and dimensionality reduction. Furthermore, we outline foundations and recent work in emerging visualization topics such as topological data analysis and understanding deep learning model internals. Index Terms–visualization, machine learning

Palavras-chave: Dimensionality reduction, Deep learning, Visualization, Analytical models, Data visualization, Machine learning, Tutorials