New guidance for using t-SNE: Alternative defaults, hyperparameter selection automation, and comparative evaluation
We present new guidelines for choosing hyperparameters for t-SNE and an evaluation comparing these guidelines to current ones. These guidelines include a proposed empirically optimum guideline derived from a t-SNE hyperparameter grid search over a large collection of data sets. We also introduce a n...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2022-06-01
|
| Series: | Visual Informatics |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2468502X22000201 |
| _version_ | 1828438673413111808 |
|---|---|
| author | Robert Gove Lucas Cadalzo Nicholas Leiby Jedediah M. Singer Alexander Zaitzeff |
| author_facet | Robert Gove Lucas Cadalzo Nicholas Leiby Jedediah M. Singer Alexander Zaitzeff |
| author_sort | Robert Gove |
| collection | DOAJ |
| description | We present new guidelines for choosing hyperparameters for t-SNE and an evaluation comparing these guidelines to current ones. These guidelines include a proposed empirically optimum guideline derived from a t-SNE hyperparameter grid search over a large collection of data sets. We also introduce a new method to featurize data sets using graph-based metrics called scagnostics; we use these features to train a neural network that predicts optimal t-SNE hyperparameters for the respective data set. This neural network has the potential to simplify the use of t-SNE by removing guesswork about which hyperparameters will produce the best embedding. We evaluate and compare our neural network-derived and empirically optimum hyperparameters to several other t-SNE hyperparameter guidelines from the literature on 68 data sets. The hyperparameters predicted by our neural network yield embeddings with similar accuracy as the best current t-SNE guidelines. Using our empirically optimum hyperparameters is simpler than following previously published guidelines but yields more accurate embeddings, in some cases by a statistically significant margin. We find that the useful ranges for t-SNE hyperparameters are narrower and include smaller values than previously reported in the literature. Importantly, we also quantify the potential for future improvements in this area: using data from a grid search of t-SNE hyperparameters we find that an optimal selection method could improve embedding accuracy by up to two percentage points over the methods examined in this paper. |
| first_indexed | 2024-12-10T20:08:32Z |
| format | Article |
| id | doaj.art-594babfbfc27491ca8f51a4869de6823 |
| institution | Directory Open Access Journal |
| issn | 2468-502X |
| language | English |
| last_indexed | 2024-12-10T20:08:32Z |
| publishDate | 2022-06-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Visual Informatics |
| spelling | doaj.art-594babfbfc27491ca8f51a4869de68232022-12-22T01:35:20ZengElsevierVisual Informatics2468-502X2022-06-01628797New guidance for using t-SNE: Alternative defaults, hyperparameter selection automation, and comparative evaluationRobert Gove0Lucas Cadalzo1Nicholas Leiby2Jedediah M. Singer3Alexander Zaitzeff4Corresponding author.; Two Six Technologies, USATwo Six Technologies, USATwo Six Technologies, USATwo Six Technologies, USATwo Six Technologies, USAWe present new guidelines for choosing hyperparameters for t-SNE and an evaluation comparing these guidelines to current ones. These guidelines include a proposed empirically optimum guideline derived from a t-SNE hyperparameter grid search over a large collection of data sets. We also introduce a new method to featurize data sets using graph-based metrics called scagnostics; we use these features to train a neural network that predicts optimal t-SNE hyperparameters for the respective data set. This neural network has the potential to simplify the use of t-SNE by removing guesswork about which hyperparameters will produce the best embedding. We evaluate and compare our neural network-derived and empirically optimum hyperparameters to several other t-SNE hyperparameter guidelines from the literature on 68 data sets. The hyperparameters predicted by our neural network yield embeddings with similar accuracy as the best current t-SNE guidelines. Using our empirically optimum hyperparameters is simpler than following previously published guidelines but yields more accurate embeddings, in some cases by a statistically significant margin. We find that the useful ranges for t-SNE hyperparameters are narrower and include smaller values than previously reported in the literature. Importantly, we also quantify the potential for future improvements in this area: using data from a grid search of t-SNE hyperparameters we find that an optimal selection method could improve embedding accuracy by up to two percentage points over the methods examined in this paper.http://www.sciencedirect.com/science/article/pii/S2468502X22000201Dimensionality reductionMachine learningt-SNE |
| spellingShingle | Robert Gove Lucas Cadalzo Nicholas Leiby Jedediah M. Singer Alexander Zaitzeff New guidance for using t-SNE: Alternative defaults, hyperparameter selection automation, and comparative evaluation Visual Informatics Dimensionality reduction Machine learning t-SNE |
| title | New guidance for using t-SNE: Alternative defaults, hyperparameter selection automation, and comparative evaluation |
| title_full | New guidance for using t-SNE: Alternative defaults, hyperparameter selection automation, and comparative evaluation |
| title_fullStr | New guidance for using t-SNE: Alternative defaults, hyperparameter selection automation, and comparative evaluation |
| title_full_unstemmed | New guidance for using t-SNE: Alternative defaults, hyperparameter selection automation, and comparative evaluation |
| title_short | New guidance for using t-SNE: Alternative defaults, hyperparameter selection automation, and comparative evaluation |
| title_sort | new guidance for using t sne alternative defaults hyperparameter selection automation and comparative evaluation |
| topic | Dimensionality reduction Machine learning t-SNE |
| url | http://www.sciencedirect.com/science/article/pii/S2468502X22000201 |
| work_keys_str_mv | AT robertgove newguidanceforusingtsnealternativedefaultshyperparameterselectionautomationandcomparativeevaluation AT lucascadalzo newguidanceforusingtsnealternativedefaultshyperparameterselectionautomationandcomparativeevaluation AT nicholasleiby newguidanceforusingtsnealternativedefaultshyperparameterselectionautomationandcomparativeevaluation AT jedediahmsinger newguidanceforusingtsnealternativedefaultshyperparameterselectionautomationandcomparativeevaluation AT alexanderzaitzeff newguidanceforusingtsnealternativedefaultshyperparameterselectionautomationandcomparativeevaluation |