Question

In the context of dimensionality reduction for single-cell data, which technique preserves local neighborhoods by minimizing the divergence between probability distributions of pairs in high and low-dimensional space?

Accepted Answer

The technique is t-distributed Stochastic Neighbor Embedding, commonly known as t-SNE. In this process, the high-dimensional data is represented as a probability distribution where the likelihood of picking a neighbor is proportional to its proximity to a point. t-SNE converts these distances into probabilities by using a Gaussian distribution for high-dimensional space and a Student’s t-distribution for low-dimensional space. The algorithm then minimizes the difference between these two probability distributions using Kullback-Leibler divergence, a mathematical measure of how one probability distribution differs from another. By minimizing this divergence, the algorithm forces points that are close together in the original high-dimensional space to remain close in the low-dimensional visualization, effectively preserving the local neighborhood structure of the single-cell data.

Home → All Courses → Engineering and Technology Courses → Bioinformatics Engineering → Flashcard

In the context of dimensionality reduction for single-cell data, which technique preserves local neighborhoods by minimizing the divergence between probability distributions of pairs in high and low-dimensional space?

The technique is t-distributed Stochastic Neighbor Embedding, commonly known as t-SNE. In this process, the high-dimensional data is represented as a probability distribution where the likelihood of picking a neighbor is proportional to its proximity to a point. t-SN....