Tempus Core: Area-Power Efficient Temporal-Unary Convolution Core for Low-Precision Edge DLAs

Autor: Vellaisamy, Prabhu, Nair, Harideep, Kang, Thomas, Ni, Yichen, Fan, Haoyang, Qi, Bin, Chen, Jeff, Blanton, Shawn, Shen, John Paul

The increasing complexity of deep neural networks (DNNs) poses significant challenges for edge inference deployment due to resource and power constraints of edge devices. Recent works on unary-based matrix multiplication hardware aim to leverage data

Externí odkaz: http://arxiv.org/abs/2412.19002

tuGEMM: Area-Power-Efficient Temporal Unary GEMM Architecture for Low-Precision Edge AI

Autor: Nair, Harideep, Vellaisamy, Prabhu, Chen, Albert, Finn, Joseph, Li, Anna, Trivedi, Manav, Shen, John Paul

General matrix multiplication (GEMM) is a ubiquitous computing kernel/algorithm for data processing in diverse applications, including artificial intelligence (AI) and deep learning (DL). Recent shift towards edge computing has inspired GEMM architec

Externí odkaz: http://arxiv.org/abs/2412.17966

Pointing Accuracy Improvements for the South Pole Telescope with Machine Learning

Autor: Chichura, P. M., Rahlin, A., Anderson, A. J., Ansarinejad, B., Archipley, M., Balkenhol, L., Benabed, K., Bender, A. N., Benson, B. A., Bianchini, F., Bleem, L. E., Bouchet, F. R., Bryant, L., Camphuis, E., Carlstrom, J. E., Chang, C. L., Chaubal, P., Chokshi, A., Chou, T. -L., Coerver, A., Crawford, T. M., Daley, C., de Haan, T., Dibert, K. R., Dobbs, M. A., Doohan, M., Doussot, A., Dutcher, D., Everett, W., Feng, C., Ferguson, K. R., Fichman, K., Foster, A., Galli, S., Gambrel, A. E., Gardner, R. W., Ge, F., Goeckner-Wald, N., Gualtieri, R., Guidi, F., Guns, S., Halverson, N. W., Hivon, E., Holder, G. P., Holzapfel, W. L., Hood, J. C., Hryciuk, A., Huang, N., Kéruzoré, F., Khalife, A. R., Kim, J., Knox, L., Korman, M., Kornoelje, K., Kuo, C. -L., Levy, K., Lowitz, A. E., Lu, C., Maniyar, A., Marrone, D. P., Martsen, E. S., Menanteau, F., Millea, M., Montgomery, J., Nakato, Y., Natoli, T., Noble, G. I., Omori, Y., Padin, S., Pan, Z., Paschos, P., Phadke, K. A., Pollak, A. W., Prabhu, K., Quan, W., Rahimi, M., Reichardt, C. L., Rouble, M., Ruhl, J. E., Schiappucci, E., Sobrin, J. A., Stark, A. A., Stephen, J., Tandoi, C., Thorne, B., Trendafilova, C., Umilta, C., Veitch-Michaelis, J., Vieira, J. D., Vitrier, A., Wan, Y., Whitehorn, N., Wu, W. L. K., Young, M. R., Zagorski, K., Zebrowski, J. A.

We present improvements to the pointing accuracy of the South Pole Telescope (SPT) using machine learning. The ability of the SPT to point accurately at the sky is limited by its structural imperfections, which are impacted by the extreme weather at

Externí odkaz: http://arxiv.org/abs/2412.15167

ONEBench to Test Them All: Sample-Level Benchmarking Over Open-Ended Capabilities

Autor: Ghosh, Adhiraj, Dziadzio, Sebastian, Prabhu, Ameya, Udandarao, Vishaal, Albanie, Samuel, Bethge, Matthias

Traditional fixed test sets fall short in evaluating open-ended capabilities of foundation models. To address this, we propose ONEBench(OpeN-Ended Benchmarking), a new testing paradigm that consolidates individual evaluation datasets into a unified,

Externí odkaz: http://arxiv.org/abs/2412.06745

Quark: Real-time, High-resolution, and General Neural View Synthesis

Autor: Flynn, John, Broxton, Michael, Murmann, Lukas, Chai, Lucy, DuVall, Matthew, Godard, Clément, Heal, Kathryn, Kaza, Srinivas, Lombardi, Stephen, Luo, Xuan, Achar, Supreeth, Prabhu, Kira, Sun, Tiancheng, Tsai, Lynn, Overbeck, Ryan

We present a novel neural algorithm for performing high-quality, high-resolution, real-time novel view synthesis. From a sparse set of input RGB images or videos streams, our network both reconstructs the 3D scene and renders novel views at 1080p res

Externí odkaz: http://arxiv.org/abs/2411.16680