Advances in Self-Supervised Learning and Robust Data Analysis

The field of data analysis and machine learning is witnessing significant developments, with a growing emphasis on self-supervised learning and robust data analysis techniques. Researchers are exploring innovative methods to improve the accuracy and efficiency of data processing, including deep subspace clustering, tensor robust principal component analysis, and blind source separation. These advances aim to address the limitations of traditional approaches, such as the need for labeled data, hyperparameter tuning, and sensitivity to outliers. Notable papers in this area include:

  • A novel single-view deep subspace clustering approach that minimizes layer-wise self-expression loss and optimizes subspace-structured norms, outperforming linear methods with carefully tuned hyperparameters.
  • A self-guided data augmentation approach for tensor robust principal component analysis, demonstrating improvements in accuracy and computational efficiency compared to state-of-the-art methods.

Sources

Label-independent hyperparameter-free self-supervised single-view deep subspace clustering

Outlier-aware Tensor Robust Principal Component Analysis with Self-guided Data Augmentation

An SE(3) Noise Model for Range-Azimuth-Elevation Sensors

The Masked Matrix Separation Problem: A First Analysis

Blind Source Separation Based on Sparsity

Provable algorithms for multi-reference alignment over $\SO(2)$

Quaternion Nuclear Norms Over Frobenius Norms Minimization for Robust Matrix Completion

Robust Orthogonal NMF with Label Propagation for Image Clustering

Built with on top of