About me       News    Publications    Talks    Contacts

About me

I am currently a Research Fellow in the Department of Neurology at Mayo Clinic, Arizona, working with Dr. Oana Dumitrascu on early detection and risk modeling of Preclinical Alzheimer's disease. Prior to that, I was a postdoc at Arizona State University, where I explored responsible machine learning methods with Dr. Visar Berisha and Dr. Gautam Dasarathy. I completed my Ph.D. at the Department of Electrical and Computer Engineering at the University of Central Florida.

News

New paper: "Unraveling overoptimism and publication bias in ML-driven science" Pouria Saidi, Gautam Dasarathy, and Visar Berisha. Published in Patterns (Cell Press), now available online.

Presented our recent findings at ITA 2023, titled "Realistic Estimates of Machine Learning Models Classification Accuracy from Overoptimistic Published Results".

Projects and Publications

Responsible Machine Learning

• Pouria Saidi, Gautam Dasarathy and Visar Berisha. Unraveling overoptimism and publication bias in ML-driven science. Patterns. 2025
• Sean Kinahan, Pouria Saidi, Ayoub Daliri, Julie Liss and Visar Berisha. Achieving Reproducibility in EEG-Based Machine Learning. FACCT 2024

Sparse recovery guarantees for Nested Periodic Dictionaries

• Pouria Saidi and George Atia. Structured Model Selection for Recovery of Sparse Representations in Nested Periodic Dictionaries. In 56th Asilomar Conference on Signals, Systems, and Computers, pages 862–866, 2022
• Pouria Saidi and George Atia. Support recovery of periodic mixtures with nested periodic dictionaries. Submitted to IEEE Journal of Selected Areas on Information Theory, preprint available arXiv:2110.13200, November 2021
• Pouria Saidi and George K Atia. Sparse recovery guarantees of periodic signals with nested periodic dictionaries. IEEE Information Theory Workshop (ITW), pages 1–6, 2021
• Pouria Saidi, George Atia, and Azadeh Vosoughi. Improved block-sparse recovery bound using cumulative block coherence. In 54th Asilomar Conference on Signals, Systems, and Computers, pages 1385–1389, 2020

Detection of visually evoked potentials to develop brain computer interfaces

• Pouria Saidi, Azadeh Vosoughi, and George Atia. Detection of brain stimuli using Ramanujan periodicity transforms. Journal of Neural Engineering, 2019
• Pouria Saidi, George Atia, and Azadeh Vosoughi. Detection of visual evoked potentials using Ramanujan periodicity transform for real time brain computer interfaces. In ICASSP, 2017
• Pouria Saidi, George Atia, and Azadeh Vosoughi. On robust detection of brain stimuli with Ramanujan periodicity transforms. In Asilomar Conference on Signals, Systems, and Computers, 2017
• Pouria Saidi, George Atia, and Azadeh Vosoughi. Periodicity transforms for multichannel and multiclass detection of visual evoked potentials. In 52nd Asilomar Conference on Signals, Systems, and Computers, 2018

Motor imagery-based brain computer interface

• Pouria Saidi, George Atia, and et al. Motor imagery classification using multiresolution analysis and sparse representation of EEG signals. In GlobalSIP, 2015

Talks

• Estimating Realistic Performance of Machine Learning Models in the Face of Publication Bias - Reproducibility, Replicability and Trust in Science, September 2022
• Realistic Estimates of Machine Learning Models Classification Accuracy from Overoptimistic Published Results, Information Theory and Applications Workshop (ITA), San Diego, February 2023
• Estimating Realistic Classification Accuracy of Machine Learning Models from Overoptimistic Published Results, Quantitative Psychology Seminar Series, Arizona State University,January 2023

Contacts

• psaidi@asu.edu
• pouria.saidi@gmail.com