Use this link to cite:
http://hdl.handle.net/2183/33866 Análisis del impacto de las medidas de distancia en técnicas de reducción de la dimensionalidad
Loading...
Identifiers
Publication date
Authors
Dorado Valín, Antía
Advisors
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen]: Debido a que la gran cantidad de datos generada por empresas, instituciones y usuarios es cada
vez mayor, múltiples veces es necesario lidiar con conjuntos de datos con elevado número
de características, lo que hace que las técnicas de reducción de la dimensión cobren cada vez
mayor importancia. En este proyecto se va a analizar el rendimiento de diferentes medidas
de distancia en técnicas de reducción de la dimensión, en concreto, en Principal Component
Analysis (PCA), t-Distributed Stochastic Neighbor Embedding (t-SNE) y Uniform Manifold Approximation
and Projection (UMAP). El objetivo es observar como afectan estas medidas de
distancia tanto en la visualización como en la clasificación del conjunto de datos. Se va a establecer
como baseline la distancia euclídea y se va a comparar con resultados obtenidos con
otras métricas como Canberra, correlación, Minkowski o coseno.
[Abstract]:Due to the large amount of data generated by companies, institutions and users, it is often necessary to deal with data sets with a high number of features, which makes dimension reduction techniques increasingly important. In this project, the performance of different distance measures in dimension reduction techniques will be analyzed, specifically, in Principal Component Analysis (PCA), t-Distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP). The objective is to observe how these distance measures affect both the visualization and the classification of the data set. The euclidean distance will be established as a baseline and it will be compared with results obtained with other metrics such as Canberra, correlation, Minkowski or cosine.
[Abstract]:Due to the large amount of data generated by companies, institutions and users, it is often necessary to deal with data sets with a high number of features, which makes dimension reduction techniques increasingly important. In this project, the performance of different distance measures in dimension reduction techniques will be analyzed, specifically, in Principal Component Analysis (PCA), t-Distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP). The objective is to observe how these distance measures affect both the visualization and the classification of the data set. The euclidean distance will be established as a baseline and it will be compared with results obtained with other metrics such as Canberra, correlation, Minkowski or cosine.
Description
Editor version
Rights
Atribución-CompartirIgual 3.0 España








