Robust Unsupervised Multi-task and Transfer Learning on Gaussian Mixture Models

Ye Tian published a research paper on multi-task and transfer learning for Gaussian Mixture Models (GMMs) on May 18, 2026. The work includes a proposed learning procedure based on the EM algorithm, theoretical guarantees, and demonstrations through simulations and real data examples. The dataset consists of supporting files in PDF and GZ formats totaling 1.5 MB.

Use Cases

• Developing robust multi-task learning algorithms based on the proposed EM procedure for GMMs.
• Studying transfer learning performance for clustering tasks using the generalized theoretical framework.
• Benchmarking initialization alignment methods for iterative unsupervised learning algorithms.
• Validating theoretical convergence rates for parameter estimation and mis-clustering errors in GMMs.

Strengths

• The work provides theoretical guarantees, including minimax optimal convergence rates for parameter estimation and excess mis-clustering error.
• The proposed method is described as robust against a fraction of outlier tasks from arbitrary distributions.
• Includes demonstrations through both simulations and real data examples.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Ye Tian via figshare

Collection Method

Likely generated for research and simulation purposes, as described in the paper.

Freshness

Last updated 2026-05-18 13:49:06; freshness should be verified.