Paper accepted at AISTATS 2026: Efficient and Accurate Tensor Compression via Recursive Sketching

Congratulations to the authors Amit Sharma, Mohammad Azhar Khan, and Dr.Rameshwar Pratap on this outstanding achievement!

We are delighted to announce that the paper titled “Efficient and Accurate Tensor Compression via Recursive Sketching” has been accepted at AISTATS 2026.

Key Highlight:

This work proposes a novel sketching algorithm that enables unbiased inner-product estimation with variance completely independent of tensor order — overcoming a major limitation of existing tensor compression methods. In addition, the approach achieves significantly faster computation, making it highly practical for large-scale tensor processing.

Heartfelt congratulations to all the authors for this excellent contribution to the research community!