Siddarth Asokan

PhD 🏅

Biography

Dr. Siddarth Asokan is currently a Senior Researcher at the Microsoft Research Lab (MSR) in Bengaluru, India. Prior to joining MSR, he received both a Ph.D. and M.Tech. (Research) Degree from the Department of Cyber Physical Systems at the Indian Institute of Science in 2023, and a Bachelor of Engineering (B.E.) degree in Electronics and Communication Engineering from the M.S. Ramaiah Institute of Technology, Bangalore in 2017. During his Ph.D., he worked on generative machine learning for images and developed strong theoretical foundations for the widely popular generative adversarial networks (GANs) and diffusion model frameworks with significant contributions to high-dimensional interpolation, Fourier approximations, and partial differential equations in very large dimensional space. His doctoral research was awarded the IUPRAI Doctoral Dissertation Award 2023, and the Prof. Satish Dhawan Research Award 2024. He has received various accolades in the past, including the IEI Young Engineers' Award 2024-25, the Qualcomm Innovation Fellowship in 2019, 2021, 2022, and 2023, the Robert Bosch Center for Cyber Physical Systems Fellowship in 2020 and 2021, and the Microsoft Research Fellowship in 2018. He also has several high-profile publications at NeurIPS, ICML, CVPR, and JMLR. He currently works at the intersection of generative modeling and large-scale information retrieval models. His research interests lie broadly in the space of signal processing, image processing, information retrieval, and generative machine learning, with a focus on building mathematically well-founded generative learning frameworks.

Publications

2024

ICASSP 24

Variational Analysis of Adversarial Regularization for Solving Inverse Problems

A. S. Bhandiwad, A. J. Kamath, S. Asokan, and C. S. Seelamantula

In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024

BIB URL

@inproceedings{bhandiwad2024variational,
  author = {Bhandiwad, A. S. and Kamath, A. J. and Asokan, S. and Seelamantula, C. S.},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  title = {Variational Analysis of Adversarial Regularization for Solving Inverse Problems},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&cstart=100&pagesize=100&citation_for_view=1g1i1B4AAAAJ:GFxP56DSvIMC},
  year = {2024}
}

ICASSP 24

Momentum-Imbued Langevin Dynamics (MILD) for Faster Sampling

N. Shetty, M. Bandla, N. Neema, S. Asokan, and C. S. Seelamantula

In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024

BIB URL

@inproceedings{shetty2024momentum,
  author = {Shetty, N. and Bandla, M. and Neema, N. and Asokan, S. and Seelamantula, C. S.},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  title = {Momentum-Imbued Langevin Dynamics (MILD) for Faster Sampling},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&cstart=200&pagesize=100&citation_for_view=1g1i1B4AAAAJ:ZzlSgRqYykMC},
  year = {2024}
}

2023

JMLR 23

Euler-Lagrange analysis of generative adversarial networks

S. Asokan, and C. S. Seelamantula

Journal of Machine Learning Research, 2023

BIB URL

@article{asokan2023euler,
  author = {Asokan, S. and Seelamantula, C. S.},
  journal = {Journal of Machine Learning Research},
  number = {126},
  pages = {1--100},
  title = {Euler-Lagrange analysis of generative adversarial networks},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&cstart=100&pagesize=100&citation_for_view=1g1i1B4AAAAJ:yMeIxYmEMEAC},
  volume = {24},
  year = {2023}
}

arXiv 23

Data Interpolants–That’s What Discriminators in Higher-order Gradient-regularized GANs Are

S. Asokan, and C. S. Seelamantula

arXiv preprint arXiv:2306.00785, 2023

BIB URL

@article{asokan2023data,
  author = {Asokan, S. and Seelamantula, C. S.},
  journal = {arXiv preprint arXiv:2306.00785},
  title = {Data Interpolants--That's What Discriminators in Higher-order Gradient-regularized GANs Are},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&cstart=100&pagesize=100&citation_for_view=1g1i1B4AAAAJ:r_AWSJRzSzQC},
  year = {2023}
}

CVPR 23

Spider GAN: Leveraging friendly neighbors to accelerate GAN training

S. Asokan, and C. S. Seelamantula

In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023

BIB URL

@inproceedings{asokan2023spider,
  author = {Asokan, S. and Seelamantula, C. S.},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  title = {Spider GAN: Leveraging friendly neighbors to accelerate GAN training},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&cstart=100&pagesize=100&citation_for_view=1g1i1B4AAAAJ:fFSKOagxvKUC},
  year = {2023}
}

arXiv 23

Gans settle scores!

S. Asokan , N. Shetty, A. Srikanth, and C. S. Seelamantula

arXiv preprint arXiv:2306.01654, 2023

BIB URL

@article{asokan2023gans,
  author = {Asokan, S. and Shetty, N. and Srikanth, A. and Seelamantula, C. S.},
  journal = {arXiv preprint arXiv:2306.01654},
  title = {Gans settle scores!},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&cstart=100&pagesize=100&citation_for_view=1g1i1B4AAAAJ:LhH-TYMQEocC},
  year = {2023}
}

ICASSP 23

A Game of Snakes and Gans

S. Asokan, F. S. Mohammed, and C. S. Seelamantula

In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023

BIB URL

@inproceedings{asokan2023game,
  author = {Asokan, S. and Mohammed, F. S. and Seelamantula, C. S.},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  title = {A Game of Snakes and Gans},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&cstart=200&pagesize=100&citation_for_view=1g1i1B4AAAAJ:OR75R8vi5nAC},
  year = {2023}
}

2022

NeurIPS-W 22

LSGANs with gradient regularizers are smooth high-dimensional interpolators

S. Asokan, and C. S. Seelamantula

In First Workshop on Interpolation Regularizers and Beyond at NeurIPS, 2022

BIB URL

@inproceedings{asokan2022lsgans,
  author = {Asokan, S. and Seelamantula, C. S.},
  booktitle = {First Workshop on Interpolation Regularizers and Beyond at NeurIPS},
  title = {LSGANs with gradient regularizers are smooth high-dimensional interpolators},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&cstart=100&pagesize=100&citation_for_view=1g1i1B4AAAAJ:mNrWkgRL2YcC},
  year = {2022}
}

DLDE-II 22

Bridging the gap between Coulomb GAN and gradient-regularized WGAN

S. Asokan, and C. S. Seelamantula

In The Symbiosis of Deep Learning and Differential Equations II, 2022

BIB URL

@inproceedings{asokan2022bridging,
  author = {Asokan, S. and Seelamantula, C. S.},
  booktitle = {The Symbiosis of Deep Learning and Differential Equations II},
  title = {Bridging the gap between Coulomb GAN and gradient-regularized WGAN},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&cstart=100&pagesize=100&citation_for_view=1g1i1B4AAAAJ:cWzG1nlazyYC},
  year = {2022}
}

2020

NeurIPS 20

Teaching a gan what not to learn

S. Asokan, and C. Seelamantula

Advances in Neural Information Processing Systems, 2020

BIB URL

@article{asokan2020teaching,
  author = {Asokan, S. and Seelamantula, C.},
  journal = {Advances in Neural Information Processing Systems},
  pages = {3964--3975},
  title = {Teaching a gan what not to learn},
  url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=1g1i1B4AAAAJ&pagesize=100&citation_for_view=1g1i1B4AAAAJ:LO7wyVUgiFcC},
  volume = {33},
  year = {2020}
}