Burcu Canakci

Analog Diffusion Models

Jiaqi Chu, H.H.C. Kremer, Fabian Falck, Grace Brennan et autres

gb (code pays fourni par la source)

Analog Diffusion Models

Jiaqi Chu, H.H.C. Kremer, Fabian Falck, Grace Brennan et autres

gb (code pays fourni par la source)

Laser writing in glass for dense, fast and efficient archival data storage

James Clegg, Patrick Anderson, Erika Aranas, Youssef Assaf et autres

Long-term preservation of digital information is vital for safeguarding the knowledge of humanity for future generations. Existing archival storage solutions, such as magnetic tapes and hard disk drives, suffer from limited media lifespans that render them unsuitable for long-term data retention1–3. Optical …

gb (code pays fourni par la source)

Analog Optical Computing: Toward Sustainable Machine Learning Models and Beyond

Francesca R. Parmigiani, Hitesh Ballani, Grace Brennan, Burcu Canakci et autres

We present a low-cost, room-temperature Analog Optical Computer (AOC) and its applications. Through hardware-applications co-design, AOC leverages the unique speed and parallelism of optics to accelerate machine learning inference and complex optimization problems by up to$100 \times$over conventional digital systems.

gb (code pays fourni par la source)

Analog optical computer for AI inference and combinatorial optimization

Kirill P. Kalinin, Jannes Gladrow, Jiaqi Chu, James Clegg et autres

target either AI or optimization workloads and rely on frequent, energy-intensive digital conversions, limiting efficiency. These systems also face application-hardware mismatches, whether handling memory-bottlenecked neural models, mapping real-world optimization problems or contending with inherent analog noise. Here we introduce an analog optical …

gb, us (code pays fourni par la source)

Accelerating ML Models and NP-Hard Optimisation Problems Using Light

Francesca R. Parmigiani, Hitesh Ballani, Grace Brennan, Burcu Canakci et autres

In this talk, I shall introduce our efforts to build an Analog Optical Computer (AOC) to speed-up specialized computations. We estimate that AOC can offer more than 100x improvement in overall system efficiency in terms of Tera Operations per Second per Watt …

gb (code pays fourni par la source)

Storage Class Memory is Dead, All Hail Managed-Retention Memory: Rethinking Memory for the AI Era

Sergey Legtchenko, Ioan A. Stefanovici, Richard T. Black, Antony I.T. Rowstron et autres

AI clusters today are one of the major uses of High Bandwidth Memory (HBM). However, HBM is suboptimal for AI workloads for several reasons. Analysis shows HBM is overprovisioned on write performance, but underprovisioned on density and read bandwidth, and also has …

gb (code pays fourni par la source)

Good things come in small packages: Should we build AI clusters with Lite-GPUs?

Burcu Canakci, Junyi Liu, Xingbo Wu, Nathanaël Cheriere et autres

To match the blooming demand of generative AI workloads, GPU designers have so far been trying to pack more and more compute and memory into single complex and expensive packages. However, there is growing uncertainty about the scalability of individual GPUs and …

gb (code pays fourni par la source)

Good things come in small packages: Should we build AI clusters with Lite-GPUs?

Burcu Canakci, Junyi Liu, Xingbo Wu, Nathanaël Cheriere et autres

To match the blooming demand of generative AI workloads, GPU designers have so far been trying to pack more and more compute and memory into single complex and expensive packages. However, there is growing uncertainty about the scalability of individual GPUs and …

Managed-Retention Memory: A New Class of Memory for the AI Era

Sergey Legtchenko, Ioan A. Stefanovici, Richard T. Black, Antony I.T. Rowstron et autres

AI clusters today are one of the major uses of High Bandwidth Memory (HBM). However, HBM is suboptimal for AI workloads for several reasons. Analysis shows HBM is overprovisioned on write performance, but underprovisioned on density and read bandwidth, and also has …

Roadmap on Neuromorphic Photonics

Daniel Brunner, Bhavin J. Shastri, Mohammed Qadasi, Hitesh Ballani et autres

This roadmap consolidates recent advances while exploring emerging applications, reflecting the remarkable diversity of hardware platforms, neuromorphic concepts, and implementation philosophies reported in the field. It emphasizes the critical role of cross-disciplinary collaboration in this rapidly evolving field.

fr, be, de, us, ca (code pays fourni par la source)

Analog Optical Computing for Sustainable AI and Beyond

Francesca R. Parmigiani, Hitesh Ballani, Grace Brennan, Burcu Canakci et autres

I will introduce the Analog Optical Computer, which, through co-design of applications and integrated 3D hardware, has the potential to accelerate by >100-fold machine learning tasks and real-world optimization problems, like medical imaging and financial transactions Full-text article not available; see video …

gb (code pays fourni par la source)