General-purpose HPO for unsupervised learning
Develop general-purpose hyperparameter optimization algorithms for unsupervised learning, where suitable response functions and performance metrics are not readily available, so that such methods can be applied across tasks like anomaly detection and generative modeling without requiring task-specific tuning procedures.
References
Still, the development of general purpose HPO algorithms for unsupervised learning remains an open problem.
— Hyperparameter Optimization in Machine Learning
(2410.22854 - Franceschi et al., 2024) in Conclusions, subsection "Response functions for unsupervised learning"
Given that identifying the optimal balance parameter \lambda is still an unresolved issue in unsupervised learning, we adopt a parameter-tuning approach.
— Reliable Neural Collapse Approximation for Open-World Test-Time Adaptation
(2608.19890 - Lin et al., 20 Aug 2026) in Section 4.8, “Parameter Analysis”
Selecting \ensuremath{\varepsilon} without labels, for example from the reconstruction error on normal validation images, remains open.
— Quasi-Binarized Autoencoders: An Architecture-Independent Information Bottleneck for Medical Image Anomaly Detection
(2610.09670 - Hanaoka et al., 7 Oct 2026) in Section 6, subsection “Limitations,” paragraph “Selection on the test set”