Median Consensus Embedding for Dimensionality Reduction
This addresses instability issues in dimensionality reduction for data analysis, but it is incremental as it builds on existing embedding methods.
The study tackled the problem of variability in low-dimensional embeddings due to random initialization in techniques like t-SNE by proposing median consensus embedding (MCE), which uses the geometric median of multiple embeddings and achieves consistency at an exponential rate, with real-world data showing rapid convergence and significant instability reduction.
This study proposes median consensus embedding (MCE) to address variability in low-dimensional embeddings caused by random initialization in dimensionality reduction techniques such as t-distributed stochastic neighbor embedding. MCE is defined as the geometric median of multiple embeddings. By assuming multiple embeddings as independent and identically distributed random samples and applying large deviation theory, we prove that MCE achieves consistency at an exponential rate. Furthermore, we develop a practical algorithm to implement MCE by constructing a distance function between embeddings based on the Frobenius norm of the pairwise distance matrix of data points. Application to real-world data demonstrates that MCE converges rapidly and significantly reduces instability. These results confirm that MCE effectively mitigates random initialization issues in embedding methods.