Published February 24, 2025 | Version 1.0
Dataset Open

Encoding Dimension vs MSE Accuracy of SOAP Spectra on MNIST: PCA, Linear, and Deep Autoencoders

Creators

  • 1. ROR icon Helsinki Metropolia University of Applied Sciences
  • 2. ROR icon Nihon University
  • 3. ROR icon Tokyo City University

Contributors

Researcher:

  • 1. ROR icon Nihon University
  • 2. ROR icon Helsinki Metropolia University of Applied Sciences

Description

This dataset contains numerical data for the plots that analyze the effect of encoding dimension on loss (MSE accuracy) for Smooth Overlap of Atomic Positions (SOAP) spectra computed on the MNIST handwritten digit dataset. The dataset includes results for three dimensionality reduction techniques:

  • Principal Component Analysis (PCA)
  • Linear Autoencoders
  • Deep Autoencoders

Files

EncodingPlotData.zip

Files (17.2 kB)

Name Size Download all
md5:56352f754791cd99e28ebf24021fefe5
17.2 kB Preview Download