2019
DOI: 10.1007/978-3-030-33778-0_20
|View full text |Cite
|
Sign up to set email alerts
|

Embedding to Reference t-SNE Space Addresses Batch Effects in Single-Cell Classification

Abstract: Dodajanje primerov v referenčno vložitev t-SNE odstrani razlike med različnimi podatkovnimi viri Tehnike zmanjševanja dimenzij, kot je t-SNE, nam omogočajo gradnjo informativnih vizualizacij visokorazsežnih naborov podatkov. Pri analizi več naborov podatkov hkrati te metode pogosto ne uspejo odkriti pomenljive skupine, temveč izpostavijo nezaželene razlike med podatkovnimi viri. Da bi odstranili vplive posameznih podatkovnih virov in odkrili strukture skupne vsem podatkom, predlagamo teoretično utemeljeno meto… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
4
0

Year Published

2020
2020
2024
2024

Publication Types

Select...
3
2
1

Relationship

0
6

Authors

Journals

citations
Cited by 7 publications
(4 citation statements)
references
References 42 publications
0
4
0
Order By: Relevance
“…We represent each reaction as a reaction Morgan difference fingerprint 39 of 2048 bits with both reagent- and non-reagent weight equal to 1. Using an implementation of parametric t-SNE from OpenTSNE, 40 we project the fingerprint vectors of reactions in USPTO 50K on the 2D plane, and then use the same t-SNE model to obtain the projections of reactions in Reaxys. The absolute values of the coordinates of the t-SNE embeddings of reaction vectors bear no physical meaning.…”
Section: Methodsmentioning
confidence: 99%
“…We represent each reaction as a reaction Morgan difference fingerprint 39 of 2048 bits with both reagent- and non-reagent weight equal to 1. Using an implementation of parametric t-SNE from OpenTSNE, 40 we project the fingerprint vectors of reactions in USPTO 50K on the 2D plane, and then use the same t-SNE model to obtain the projections of reactions in Reaxys. The absolute values of the coordinates of the t-SNE embeddings of reaction vectors bear no physical meaning.…”
Section: Methodsmentioning
confidence: 99%
“…Sequence space projections: To visualise sequence space of our aptamer selection experiments, we chose to reduce the dimensionality and project it into two dimensions using t‐Stochastic Neighbour Embedding (t‐SNE) [22] implemented in the openTSNE package v.0.4.0 [37] . Here, we used a perplexity of 50, a cosine metric, and principal component analysis for initialisation, with all other parameters as default.…”
Section: Methodsmentioning
confidence: 99%
“…This descriptor uniquely identifies crystal structures and is a continuous metric, meaning that the distance (measured using the Chebyshev metric) is zero for similar crystals. We embed the AMD to two dimensions using t-SNE, implented in openTSNE, 30 using the Chebyshev distance metric and nearest neighbour descent. 31 Fig.…”
Section: The Matgfn-rm Datasetmentioning
confidence: 99%