How to Use t-SNE Effectively¶
Wattenberg, M., Viégas, F., & Johnson, I. (2016). How to Use t-SNE Effectively. Distill, 1(10).
Cited by¶
3 citations across 3 artifacts.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Topographic Map
- Because magnification is non-uniform, between-cluster distances in a t-SNE plot are not faithful — the layout sacrifices global geometry to preserve neighbourhoods, exactly as the Mercator projection sacrifices area to preserve angles — so reading "cluster A is twice as far from B as from C" off the plot is the magnification-mismatch error the frame warns against.
This sourceDemonstrates that between-cluster distances and sizes in t-SNE plots are not faithful, the magnification-mismatch caution.
- Because magnification is non-uniform, between-cluster distances in a t-SNE plot are not faithful — the layout sacrifices global geometry to preserve neighbourhoods, exactly as the Mercator projection sacrifices area to preserve angles — so reading "cluster A is twice as far from B as from C" off the plot is the magnification-mismatch error the frame warns against.
Mechanisms¶
- Manifold / Embedding Validation
- Then the geometry checks: trustworthiness and continuity scores show local neighborhoods are only partly preserved, and a global-distance comparison shows the between-island gaps in the picture are essentially arbitrary — t-SNE inflates and rearranges global distances by design.
This sourceDemonstrates that inter-cluster distances in t-SNE plots can be misleading or meaningless.
- Then the geometry checks: trustworthiness and continuity scores show local neighborhoods are only partly preserved, and a global-distance comparison shows the between-island gaps in the picture are essentially arbitrary — t-SNE inflates and rearranges global distances by design.
- Prototype Embedding Map
- Its central hazard is that dimensionality reduction lies convincingly: cluster sizes, the distances between clusters, and even whether clusters exist at all can be artifacts of the projection's settings rather than facts about the data.
This sourceDemonstrates that t-SNE settings can distort apparent cluster size, distance, and even the presence of structure.
- Its central hazard is that dimensionality reduction lies convincingly: cluster sizes, the distances between clusters, and even whether clusters exist at all can be artifacts of the projection's settings rather than facts about the data.
Verification¶
This reference passed the adversarial substantiation pipeline: it was checked to exist and to support the claim it is attached to. See how references were verified.
Links previously used in the corpus¶
Before the registry existed this work was also linked 1 other way.
Registry ID ref:2dcdcf2f5eaf · see in the full table