-
Notifications
You must be signed in to change notification settings - Fork 1
/
info.json
26 lines (26 loc) · 1.74 KB
/
info.json
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
{
"abstract": "We propose a new tool for visualizing complex, and potentially large and high-dimensional, data sets called Centroid-Encoder (CE). The architecture of the Centroid-Encoder is similar to the autoencoder neural network but it has a modified target, i.e., the class centroid in the ambient space. As such, CE incorporates label information and performs a supervised data visualization. The training of CE is done in the usual way with a training set whose parameters are tuned using a validation set. The evaluation of the resulting CE visualization is performed on a sequestered test set where the generalization of the model is assessed both visually and quantitatively. We present a detailed comparative analysis of the method using a wide variety of data sets and techniques, both supervised and unsupervised, including NCA, non-linear NCA, t-distributed NCA, t-distributed MCML, supervised UMAP, supervised PCA, Colored Maximum Variance Unfolding, supervised Isomap, Parametric Embedding, supervised Neighbor Retrieval Visualizer, and Multiple Relational Embedding. An analysis of variance using PCA demonstrates that a non-linear preprocessing by the CE transformation of the data captures more variance than PCA by dimension.",
"authors": [
"Tomojit Ghosh",
"Michael Kirby"
],
"emails": [
"tomojit.ghosh@colostate.edu",
"Kirby@math.colostate.edu"
],
"extra_links": [
[
"code",
"https://github.com/Tomojit1/Centroid-encoder/tree/master/GPU"
]
],
"id": "20-188",
"issue": 20,
"pages": [
1,
34
],
"title": "Supervised Dimensionality Reduction and Visualization using Centroid-Encoder",
"volume": 23,
"year": 2022
}