Xiaolin Huang from Shanghai Jiao Tong University presented a talk at MBZUAI on training deep neural networks in tiny subspaces. The talk covered the low-dimension hypothesis in neural networks and methods to find subspaces for efficient training. It suggests that training in smaller subspaces can improve training efficiency, generalization, and robustness. Why it matters: Investigating efficient training methods is crucial for resource-constrained environments and can enable broader access to advanced AI.
Communications Physics journal has a focus collection on space quantum communications. The collection covers supporting technologies, new quantum protocols, inter-satellite QKD, constellations of satellites, and quantum inspired technologies and protocols for space based communication. Contributions are welcome from October 20, 2020 to April 30, 2021, and accepted papers are published on a rolling basis. Why it matters: Space-based quantum communication is a critical area for developing secure, global quantum networks, and this collection could highlight relevant research for the GCC region as it invests in advanced technologies.
Hassan Sajjad from Dalhousie University presented research on exploring the latent space of AI models to assess their safety and trustworthiness. He discussed use cases where analyzing latent space helps understand the robustness-generalization tradeoff in adversarial training and evaluate language comprehension. Sajjad's work aims to build better AI models and increase trust in their capabilities by looking at model internals. Why it matters: Intrinsic evaluation of model internals will become important to improving AI safety and robustness.
Marcus Engsig from DERC will present a paper at the MATLAB User Group Meeting in Abu Dhabi on October 6. The paper, titled ‘Generalization of Higher Order Methods For Fast Iterative Matrix Inversion Compatible With GPU Acceleration’, discusses a novel approach to matrix inversion using GPUs. The method, named Nested Neumann, achieves 4-100x acceleration compared to standard MATLAB methods for large matrices. Why it matters: This research contributes to faster computation in numerical and physical modeling, crucial for processing large datasets in various scientific and engineering applications in the region.
A cryptanalysis team at the UAE's Cryptography Research Center (CRC) has set new records in computation by decrypting a McEliece ciphertext without the secret key at INRIA’s McEliece decoding challenge, taking first and second place. The record computation took about 31.4 days on a cluster using 256 CPU-cores. The team also achieved top ranks in decoding quasi-cyclic codes and ternary codes, used in post-quantum cryptography. Why it matters: This achievement demonstrates the UAE's growing capabilities in advanced cryptography research and its contributions to the global effort to develop quantum-resistant algorithms.
The Maker Space self-directed group at KAUST promotes DIY culture and provides training on using machines, tools, and materials. In March 2017, Maker Space launched the "Design for KAUST" workshop in collaboration with the University’s Residential Maintenance Department. The winning teams in the workshop received sponsorship, including a total of SAR 10,000 in prizes, a Local Impact Award and an opportunity to test the prototypes in the field. Why it matters: This initiative fosters innovation and problem-solving within the KAUST community, addressing practical challenges in daily life through technology and promoting local impact.
MBZUAI researchers have introduced SURPRISE3D, a benchmark for evaluating 3D spatial reasoning in AI systems, along with a 3D Spatial Reasoning Segmentation (3D-SRS) task. The benchmark includes over 900 indoor scenes and 200,000 language queries paired with 3D masks, emphasizing spatial relationships over object naming. A companion paper, MLLM-For3D, explores adapting 2D multimodal LLMs for 3D reasoning. Why it matters: This work addresses a key limitation in current AI, pushing towards embodied AI that can understand and act in 3D environments based on human-like spatial reasoning.
KAUST researchers developed a machine learning algorithm to control a deformable mirror within the Subaru Telescope's exoplanet imaging camera, compensating for atmospheric turbulence. The algorithm, which computes a partial singular value decomposition (SVD), outperforms a standard SVD by a factor of four. The KAUST team received a best paper award at the PASC Conference for this work, which has already been deployed at the Subaru Telescope. Why it matters: This advancement enables sharper images of exoplanets, facilitating their identification and study, and showcases the impact of optimizing core linear algebra algorithms.