Call For Papers

Background and Aim

With the aim of automatically locating machine learning configurations with the best performance possible in constrained resources and without human involvement, automated machine learning (AutoML) views the machine learning task as a configuration search problem (i.e., optimization problem). This approach reduces the machine's reliance on human experts by enabling automatic feature processing, algorithm model selection and modeling, parameter tuning, and other tasks. AutoML has demonstrated to either match or surpass the results of human experts manually fine-tuning parameters in numerous domains, and it can significantly lower the expenses associated with implementing and utilizing machine learning. It has become one of the most popular and cutting-edge research directions in the field of artificial intelligence and machine learning.

Early AutoML are primarily driven by a single objective. However, machine learning applications in real-world naturally have more than one objective. For example, different prediction measures may be conflict objectives. In addition to prediction measures, some other potential objectives are model complexity, time consumption, and power consumption. When designing AutoML, considering multiple objectives are herein much important.

This special session aims to provide a forum for researchers and practitioners to present advanced studies in AutoML, especially the MO-AutoML.

Scope

The topics include, but are not limited to:

• Advanced techniques and applications on AutoML,

• Multiobjective automated feature engineering, such as feature selection, feature extraction, and feature,

• Multiobjective automated Neural Architecture Search (NAS),

• Multiobjective automated hyperparameter optimization,

• Multiobjective automated data augmentation,

• Multiobjective automated data cleaning,

• Multiobjective model compression,

• Multiobjective model combination,

• Multiobjective ensemble learning,

• MO-AutoML framework,

• MO-AutoML for different learning tasks, e.g., classification/regression, unsupervised, semi-supervised, self-supervised, few-shot, transfer learning,

• Real-world applications with AutoML and MO-AutoML.