Distributed Mean Estimation (DME), in which $ n $ clients communicate vectors to a
parameter server that estimates their average, is a fundamental building block in …

As deep neural networks (DNNs) grow in complexity and size, the resultant increase in
communication overhead during distributed training has become a significant bottleneck …

Quantization is a fundamental optimization for many machine-learning use cases, including
compressing gradients, model weights and activations, and datasets. The most accurate …

Gradient aggregation has long been identified as a major bottleneck in today's large-scale
distributed machine learning training systems. One promising solution to mitigate such …

In large-language models, memory constraints in the key-value cache (KVC) pose a
challenge during inference, especially with long prompts. In this work, we observed that …

Distributed Mean Estimation (DME), in which $ n $ clients communicate vectors to a
parameter server that estimates their average, is a fundamental building block in …

Software-defined networking enables tight integration between packet-processing hardware
and centralized controllers, highlighting the importance of deep network insight for informed …

Abstract Machine learning (ML) has become ubiquitous, integrating into numerous real-life
applications. However, meeting the computational demands of ML systems is challenging …