Machine Learning from Challenging Data 2025

Modern technology depends on the ubiquitous collection of data and the application of machine learning to derive insights and create knowledge. Often, machine learning methods are developed considering curated, well-behaved datasets. However, real-world data are often collected in non-ideal conditions, with limited sensing, storage, processing, and labeling capabilities, environmental changes and interference, attacks, and policy restrictions. Accordingly, real-world data presents significant challenges, such as corruptions, outliers, missing entries or labels, bias, distribution shifts, and security/privacy issues, to name just a few. Such challenges often limit the effectiveness of standard machine learning methods to real-world scenarios. The Machine Learning from Challenging Data Conference (MLCD) aims to bridge this gap by advancing practical, efficient, and effective machine learning solutions tailored to complex real-world data challenges.

We invite submissions on machine learning from challenging data. Contributors are encouraged to present highly novel methods, theoretical advancements, strategies for data collection and dataset optimization, and significant applications that demonstrate practical solutions to the complexities of real-world data scenarios.

Examples of data challenges:

• corrupted, noisy, and erroneous data
• adversarial data and security attacks
• biased and imbalanced data
• dynamic data distributions
• limited data and overfitting
• heterogenous data
• streaming data
• distributed and cloud data
• privacy-sensitive data
• multimodal data
• sparse data and missing labels
• big and high-dimensional data.

Examples of methodologies:

• robust machine learning
• continual learning
• domain adaptation
• distributed and federated learning
• transfer learning
• data synthesis and augmentation
• data curation
• optimization of sensor configuration and deployment
• dynamic and active sensing.

Examples of application areas:

• healthcare and medicine
• remote sensing
• vision
• wireless communications
• energy and power systems
• large language models
• AI and robotics
• connected AI autonomy
• IoT, ocean IoT, etc.
• sensing in extreme environments.