Dataset Distancing Lab

Compute distances between datasets using RAW, UMAP, or LWM embeddings.
Upload your .pt/.p datasets or try the built-in samples under data/{task}/{scenario}/....

Format: each file should be a Torch file with keys:

  • channels: Tensor[N, ...] (complex supported; real+imag will be concatenated)
  • labels (optional): Tensor[N]

If you use this lab or methods in your work, please cite:

@INPROCEEDINGS{10942657,
author={Morais, João and Alikhani, Sadjad and Malhotra, Akshay and Hamidi-Rad, Shahab and Alkhateeb, Ahmed},
booktitle={2024 58th Asilomar Conference on Signals, Systems, and Computers}, 
title={A Dataset Similarity Evaluation Framework for Wireless Communications and Sensing}, 
year={2024},
volume={},
number={},
pages={1144-1149},
keywords={Wireless communication;Dimensionality reduction;Adaptation models;Wireless sensor networks;Nearest neighbor methods;Extraterrestrial measurements;Data structures;Distance measurement;Data models;Sensors},
doi={10.1109/IEEECONF60004.2024.10942657}}

Framework & Distance

Framework
Distance Mode
Label weighting
8 256
32 4096

UMAP (only if Framework=UMAP)

UMAP Mode
Channel representation
4 128
2 256
2 128
0 0.99
metric
0.1 5
0.1 10
1 50
init
0 1
1 10
0.1 5

Demo datasets (data/{task}/{scenario}/...)

Task
Scenarios (files inside each scenario)

Or upload your own

Distance Matrix (Table)

Distances

Distance Matrix (Heatmap)