Algorithmic Bias Control in Deep learning

Project Information

A-B-C-Deep

Grant agreement ID: 101039436

DOI 10.3030/101039436

EC signature date 17 May 2022

Start date 1 June 2022

End date 31 May 2027

Funded under

• European Research Council (ERC)

Total cost € 1 500 000,00

EU contribution € 1 500 000,00

Coordinated by

TECHNION - ISRAEL INSTITUTE OF TECHNOLOGY
Israel

Publications

Conference proceedings (13)

How do Minimum-Norm Shallow Denoisers Look in Function Space?

Author(s): Zeno, Chen; Ongie, Greg; Blumenfeld, Yaniv; Weinberger, Nir; Soudry, Daniel
Published in: NeurIPS, 2023
Publisher: NeurIPS
DOI: 10.48550/arxiv.2311.06748

The Role of Codeword-to-Class Assignments in Error-Correcting Codes: An Empirical Study

Author(s): Evron, Itay; Onn, Ophir; Orzech, Tamar Weiss; Azeroual, Hai; Soudry, Daniel
Published in: AISTAT, 2023
Publisher: AISTAT
DOI: 10.48550/arxiv.2302.05334

DropCompute: simple and more robust distributed synchronous training via compute variance reduction

Author(s): Giladi, Niv; Gottlieb, Shahar; Shkolnik, Moran; Karnieli, Asaf; Banner, Ron; Hoffer, Elad; Levy, Kfir Yehuda; Soudry, Daniel
Published in: NeurIPS, 2023
Publisher: NeurIPS
DOI: 10.48550/arxiv.2306.10598

Explore to Generalize in Zero-Shot RL

Author(s): Zisselman, Ev; Lavie, Itai; Soudry, Daniel; Tamar, Aviv
Published in: NeurIPS, 2023
Publisher: NeurIPS
DOI: 10.48550/arxiv.2306.03072

The Implicit Bias of Minima Stability in Multivariate Shallow ReLU Networks

Author(s): Nacson, Mor Shpigel; Mulayoff, Rotem; Ongie, Greg; Michaeli, Tomer; Soudry, Daniel
Published in: ICLR, 2023
Publisher: ICLR

Continual Learning in Linear Classification on Separable Data

Author(s): Evron, Itay; Moroshko, Edward; Buzaglo, Gon; Khriesh, Maroun; Marjieh, Badea; Srebro, Nathan; Soudry, Daniel
Published in: ICML, 2023
Publisher: ICML
DOI: 10.48550/arxiv.2306.03534

Towards Cheaper Inference in Deep Networks with Lower Bit-Width Accumulators

Author(s): Blumenfeld, Yaniv; Hubara, Itay; Soudry, Daniel
Published in: ICLR, 2024
Publisher: ICLR

Gradient Descent Monotonically Decreases the Sharpness of Gradient Flow Solutions in Scalar Networks and Beyond

Author(s): Kreisler, Itai; Nacson, Mor Shpigel; Soudry, Daniel; Carmon, Yair
Published in: ICML, 2023
Publisher: ICML
DOI: 10.48550/arxiv.2305.13064

Accurate Neural Training with 4-bit Matrix Multiplications at Standard Formats

Author(s): B. Chmiel, R. Banner, E. Hoffer, H. Ben Yaacov, D. Soudry
Published in: ICLR, 2023
Publisher: ICLR

The Joint Effect of Task Similarity and Overparameterization on Catastrophic Forgetting -- An Analytical Model

Author(s): Goldfarb, Daniel; Evron, Itay; Weinberger, Nir; Soudry, Daniel; Hand, Paul
Published in: ICLR, 2024
Publisher: ICLR
DOI: 10.48550/arxiv.2401.12617

Minimum Variance Unbiased N:M Sparsity for the Neural Gradients

Author(s): B. Chmiel, I. Hubara, R. Banner, D. Soudry
Published in: ICLR, 2023
Publisher: ICLR

Alias-Free Convnets: Fractional Shift Invariance via Polynomial Activations

Author(s): Michaeli, Hagay; Michaeli, Tomer; Soudry, Daniel
Published in: CVPR, 2023
Publisher: CVPR
DOI: 10.48550/arxiv.2303.08085

How Uniform Random Weights Induce Non-uniform Bias: Typical Interpolating Neural Networks Generalize with Narrow Teachers

Author(s): G. Buzaglo*, I. Harel*, M. Shpigel Nacson*, A. Brutzkus, N. Srebro, D. Soudry
Published in: ICML, 2024
Publisher: ICML