Computer Science

SC26 connects researchers and practitioners developing high-performance computing, networking, storage and scientific data analysis. The programme includes research presentations, tutorials and workshops at McCormick Place in Chicago, 15–20 November 2026. Registration is open, with early rates through 14 October. A digital participation option is available through the official registration page.

Researchers and practitioners will discuss mathematical foundations of data science, including high-dimensional geometry, dimensionality reduction, scalable algorithms, uncertainty and machine learning. The conference takes place in person at the Salt Palace Convention Center on 16–20 November 2026. Registration is open, with early rates through 19 October; abstract and travel-support deadlines have passed. Registration also covers the co-located SIAM Imaging Science and Data Mining conferences.

The Fortieth Annual Conference on Neural Information Processing Systems brings together interdisciplinary machine-learning research through peer-reviewed sessions, invited talks, demonstrations, tutorials, workshops, and an exposition.

Katarzyna Swirydowicz uses an asynchronous solver on an AMD system to examine the practical implementation of computational linear algebra across CPUs and GPUs. The talk introduces the relevant computational ideas, programming models, and software tools, then considers how algorithmic structure interacts with hardware capabilities. This case study illustrates both the opportunities and implementation challenges of asynchronous methods for large-scale scientific computing on GPU-accelerated systems.

Graph matching algorithms attempt to find the best correspondence between the nodes of two networks. These techniques have been used to match individual neurons in nanoscale connectomes -- in particular, to find pairings of neurons across hemispheres. However, since graph matching techniques deal with two isolated networks, they have only utilized the ipsilateral (same hemisphere) subgraphs when performing the matching. Here, we present a modification to a state-of-the-art graph matching algorithm which allows it to solve what we call the bisected graph matching problem. This modification allows us to leverage the connections between the brain hemispheres when predicting neuron pairs. Via simulations and experiments on real connectome datasets, we show that this approach improves matching accuracy when sufficient edge correlation is present between the contralateral (between hemisphere) subgraphs. We also show how matching accuracy can be further improved by combining our approach with previously proposed extensions to graph matching, which utilize edge types and previously known neuron pairings. We expect that our proposed method will improve future endeavors to accurately match neurons across hemispheres in connectomes, and be useful in other applications where the bisected graph matching problem arises.

Recorded live at the International Congress of Mathematicians in Philadelphia, this episode brings hosts Janna Levin and Steven Strogatz together with Akshay Venkatesh, Ravi Vakil and Alex Kontorovich. They examine what recent AI-generated proofs demonstrate, how machine assistance may change mathematical research, and what proof, creativity, understanding and storytelling contribute to the discipline. The recording took place on 26 July 2026; Quanta published the episode on 3 September 2026.

Supports research-based graduate degrees with three funded years over a five-year fellowship: an annual $37,000 stipend plus a $16,000 institutional education allowance, subject to funding. The displayed deadline is the first subject deadline: Life Sciences on 19 October 2026. Computing, materials, psychology and social sciences are due 20 October; engineering 22 October; chemistry, geosciences, mathematics, physics and astronomy 23 October. All applications are due at 8 pm Eastern Time. Reference letters are due 16 October at 8 pm Eastern Time.

Supports research on agentic, self-improving systems that generate and optimize kernels and end-to-end workloads on AWS Trainium. Topics include generative AI for kernel optimization, performance debugging and profiling, correctness and integrity evaluation, and supervised or reinforcement learning of open-source models for kernel and system optimization. Proposed research should connect generation, evaluation and learning loops and explain plans for open-source contributions. Use the ARA proposal template; four pages excluding appendices are encouraged. Decisions are expected in February 2027. Applications close on 4 November 2026 at 11:59pm Pacific Time.

Fast computation often trades statistical accuracy for speed. In nearest-neighbor prediction, achieving optimal accuracy generally requires the number of neighbors to grow as a root of the sample size, creating a corresponding computational burden even with fast search. This talk shows how bias or variance corrections after data quantization or subsampling, together with black-box fast-search techniques, can retain accuracy while reducing prediction-time computation to O(log n). The analysis explains how much quantization or subsampling is compatible with optimal accuracy. Extensive experiments on large datasets from several domains test these theoretical insights. The talk draws on work with N. Verma and L. Xue.