Скачать или смотреть KDD2024 - Introduction to AutoXPCR: Automated Multi-Objective Model Selection for Time Series

Raphael Fischer

Introducing AutoXPCR - a novel method for automatically selecting deep learning models for time series forecasting in a resource-aware and explainable way!

Join Raphael Fischer as he explains the key concepts and take-aways from his work "AutoXPCR: Automated Multi-Objective Model Selection for Time Series Forecasting," co-authored with Amal Saadallah, which they will present at ACM KDD 2024. In this video, discover how AutoXPCR leverages meta-learning to estimate model performance in terms of Predictive error, Complexity, and Resource demand. This methodology enables to consider use case preferences and offer explanations during model selection. The video also showcases how AutoXPCR is competitive with other AutoML approaches at significantly reduced computation costs.

If this presentation raised your interest, we invite you to check out our code and preprint at https://github.com/raphischer/xpcr. Thank you for your attention, we are looking forward to meet and exchange in Barcelona, at KDD 2024!