Search papers, labs, and topics across Lattice.
This study investigates the impact of modifying the affinity matrix in t-SNE through a row-wise power transform parameterized by gamma, which allows for controlled smoothing or sharpening of pairwise similarities. The findings reveal that while sharpening enhances the preservation of nearest neighbors, smoothing is more effective for maintaining broader local neighborhood structures, outperforming existing multiscale methods. This nuanced approach to affinity matrix manipulation provides a deeper understanding of how local neighborhood preservation can be optimized in high-dimensional data visualization.
Sharpening the affinity matrix in t-SNE can significantly enhance the preservation of nearest neighbors, while smoothing broadens local neighborhood fidelity鈥攐utperforming traditional multiscale methods.
Dimensionality reduction methods are instrumental to visualize high-dimensional data, and t-SNE stands as one of the most widely used methods due to its emphasis on local neighborhood preservation. A central component of t-SNE is the affinity matrix, which expresses pairwise similarities in the form of symmetrized probabilities, over which the optimization problem of t-SNE is defined. We study how the sharpness of this probability distribution affects neighborhood preservation at different scales. We introduce a row-wise power transform controlled by a parameter gamma that can smooth or sharpen each row of the affinity matrix while preserving sparsity and rank order. We show that this transform is equivalent to rescaling the Gaussian bandwidth and thus to changing the perplexity. However, as the sharpness of the probability distribution varies per point, a fixed gamma leads to point-dependent effective perplexities, making it distinct from changing the global perplexity. Empirically, we find that sharpening improves preservation of the very nearest neighbors, while smoothing improves preservation of broader local neighborhoods, outperforming alternative affinity constructions including multiscale methods in the mid-local range.
