diemsim

diemsim is an optimized Python library to compute "Dimension Insensitive Euclidean Metric (DIEM)", surpassing Cosine similarity for multidimensional comparisons.

Latency Benchmarking

Our proposed approaches,
Compact Vectorization optimizes latency of the existing function 'DIEM_Stat' by around 46.50%
Compact Optimized getDIEM optimizes latency of the existing function 'getDIEM' by 34.27%

Getting Started

Install the package via pip:

pip install diemsim

Usage

from diemsim import DIEM

N= 12
maxV= 1
minV= 0
n_iter= int(1e5)

S1= np.random.rand(N) * (maxV - minV) + minV
S2= np.random.rand(N) * (maxV - minV) + minV

# Initialize DIEM
diem= DIEM( N= N, maxV= maxV, minV= minV, n_iter= n_iter ) 

# Compute DIEM value
value= diem.sim( S1, S2)

print( "Output Value: ", value )

Find Quick Start notebook here

Key Contributors

Boddu Sri Pavan , Chandrasheker Thummanagoti

Please refer CONTRIBUTING.md for contributions to diemsim

To cite our Python library

BibTeX

@software{diemsim,
title = {diemsim: A Python Library Implementing Dimension Insensitive Euclidean Metric (DIEM)},
author = {Boddu Sri Pavan, Chandrasheker Thummanagoti},
year = {2025},
publisher = {Zenodo},
version = {v1.0.0},
doi = {10.5281/zenodo.15351274},
url = {https://doi.org/10.5281/zenodo.15351274}
}

APA

BodduSriPavan111. (2025). BodduSriPavan-111/diemsim: Initial Release (v0.0.1). Zenodo. https://doi.org/10.5281/zenodo.15351275

Acknowledgements

BibTeX

@misc{tessari2025surpassingcosinesimilaritymultidimensional,
title={Surpassing Cosine Similarity for Multidimensional Comparisons: Dimension Insensitive Euclidean Metric},
author={Federico Tessari and Kunpeng Yao and Neville Hogan},
year={2025},
eprint={2407.08623},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2407.08623},
}

Thank You !