embedding plots

This section contains plots of embeddings for manga. Each plot is a 2D projection of the manga embedding space. The colors represent the most common tag across the neighbors of the manga. See the manga page for more information. See the manga explore page for recommendations.

descriptions

model description
lsi latent semantic indexing, low rank approximation of the tag-manga matrix, cosine distance
word2vec word2vec, neural network word embeddings on tags as a bag of words, cosine distance
network-cosine manga-manga network through tags, graph laplacian, cosine distance (questionable interpretation)
network-euclidean manga-manga network through tags, graph laplacian, euclidean distance (commute-time distance)
network-adj-cosine manga-manga network through tags, adjacency matrix, cosine distance (structural equivalence)

options

lsi

plot of Isekai in lsi using UMAP

word2vec

plot of Isekai in word2vec using UMAP

network-euclidean

plot of Isekai in network-euclidean using UMAP

network-cosine

plot of Isekai in network-cosine using UMAP

network-adj-cosine

plot of Isekai in network-adj-cosine using UMAP