]>
Word embeddings
Dimensionality reduction
Data visualisation
nonlinear dimensionality reduction technique that is particularly well suited for embedding high-dimensional data into a space of two or three dimensions, which can then be visualized in a scatter plot. Specifically, it models each high-dimensional object by a two- or three-dimensional point in such a way that similar objects are modeled by nearby points and dissimilar objects are modeled by distant points.
t-SNE
Recurrent neural network
NLP
Good
Deep Learning
Christopher Olah
2015-10-16
Deep Learning, NLP, and Representations - colah's blog