Activity and Progress

Recent events in my timeline

04.2026
Paper in ICLR-2026 at Rio de Janeiro: “Deterministic Bounds and Random Estimates of Metric Tensors on Neuromanifolds”. (TLDR: efficient unbiased FIM estimator with bounded variance, scalable to models like RoBERTa-base.) (openreview) (arxiv) (code) (slides)
04.2026
Area Chair for NeurIPS-2026
04.2026
Elevated to IEEE Senior Member.
02.2026
Expert interview: UN Women Gender Diversity in STEM & AI study
12.2025
Paper in AAAI-26 Special Track on AI for Social Impact: “Leveraging Sparse Observations to Predict Species Abundance Across Space and Time”. A collaboration with (former) CSIRO colleagues.
12.2025
AC for ICLR-2026.
09.2025
New preprint “Unbiased Online Curvature Approximation for Regularized Graph Continual Learning” Jie Yin, Ke Sun, Han Wu, 2025.
06.2025
Reviewing for GSI (Geometric Science of Information) 2025 — the biennial flagship conference in Information Geometry, with involvement (as author/reviewer) dating back to 2017.
06.2025
New article published: “A geometric modeling of Occam’s razor in deep learning”, Sun and Nielsen, Information Geometry, Special Issue: Half a Century of Information Geometry, Part 2, Springer, 2025. (arxiv) (open access link) CC BY 4.0
05.2025
My preprint “Deterministic Bounds and Random Estimates of Metric Tensors on Neuromanifolds” provides a computationally efficient method to estimate the Fisher information for deep neural networks, with a new analysis.
05.2025
Serving as an area chair of NeurIPS 2025.
10.2024
“Tradeoffs of Diagonal Fisher Information Matrix Estimators”, Soen and Sun. NeurIPS 2024 (to appear).
10.2024
Serving as an Area Chair (AC) for ICLR 2025.
07.2024
”Information Theory in Emerging Machine Learning Techniques”, a special issue of Entropy (MDPI) where I serve as the guest editor, is open for submission (deadline March 2025).
02.2024
“Tradeoffs of Diagonal Fisher Information Matrix Estimators”, Soen and Sun. Our recent follow-up of “On the Variance of the Fisher Information for Deep Learning” at NeurIPS 2021.
01.2024
Lecturing at AMSI summer school 2024.
01.2024
Serving as an Area Chair in ICLR 2024.
12.2023
Updated article “A Geometric Modeling of Occam’s Razor in Deep Learning”, Sun and Nielsen. Interested readers are invited to follow it on ResearchGate.
07.2022
Outstanding reviewer (Top 10%) of ICML 2022.
07.2022
“Secure Quantized Training for Deep Learning”, Keller and Sun, ICML, 2022. (proceedings) (arxiv)
07.2022
One of the “top reviewers” for UAI 2022.
03.2022
“Local Measurements of Non-linear Embeddings with Information Geometry”, Geometry and Statistics, Vol. 46 of Handbook of Statistics, Ed. Frank Nielsen, Arni Srinivasa Rao, C.R. Rao, 2022. (book) (arxiv)
02.2022
One of the “top reviewers” for AISTATS 2022
12.2020
Keynote “Information Geometry for Data Geometry through Pullbacks” (video) (slides) in Neurips 2020 workshop on Deep Learning through Information Geometry
11.2020
Book chapter: “Chain Rule Optimal Transport” (arxiv) in Progress in Information Geometry Theory and Applications, F. Nielsen Ed., Springer, 2021.
2020:
D61 postdoc openning (deadline soon!) on Machine Learning and Artificial Intelligence: Decision Making
2020:
TNNLS article: q-Neurons: Neuron Activations Based on Stochastic Jackson’s Derivative Operators
2020:
PC service in early 2020: ICML, IJCAI-PRICAI, UAI
2019:
You are welcome to visit our poster “Information Geometric Set Embeddings (IGSE): From Sets to Distributions” in the NeurIPS-19 workshop of Sets and Partitions.
2019:
I served as a reviewer/PC member of ICLR (outstanding reviewer), AISTATS, ICML, IJCAI, GSI, UAI, NeurIPS (one of the “top reviewers”) and did emergency reviews for ICCV, AAAI, etc.
2019:
Check our latest work on Lightlike Neuromanifolds, Occam’s Razor and Deep Learning which explains generalization of deep learning using geometric tools from general relativity.
2019:
I am attending UAI (July 22-25 2019) to present our recent work on Fisher-Bures Adversary Graph Convolutional Networks.