Communications of the ACM
Latest Research
From Communications of the ACM
Worst-Case Topological Entropy and Minimal Data Rate for State Estimation of Switched Linear Systems
In this paper, we study the problem of estimating the state of a switched linear system when the observation of the system is subject to communication constraints...
From Communications of the ACM
Technical Perspective: Neural Radiance Fields Explode on the Scene
Neural volume rendering exploded onto the scene in 2020, triggered by "NeRF," the impressive paper by Ben Mildenhall et al., on Neural Radiance Fields.
From Communications of the ACM
NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
We present a method that achieves state-of-the-art results for synthesizing novel views of complex scenes by optimizing an underlying continuous volumetric scene...
From Communications of the ACM
Technical Perspective: The Importance of WINOGRANDE
"WINOGRANDE" explores new methods of dataset development and adversarial filtering, expressly designed to prevent AI systems from making claims of smashing through...
From Communications of the ACM
WinoGrande: An Adversarial Winograd Schema Challenge at Scale
We introduce WinoGrande, a large-scale dataset of 44k problems, inspired by the original Winograd Schema Challenge, but adjusted to improve both the scale and the...
From Communications of the ACM
Technical Perspective: The Quest for Optimal Multi-Item Auctions
"Optimal Auctions Through Deep Learning," by Paul Dütting et al., contributes a very interesting and forward-looking new take on the optimal multi-item mechanism...
From Communications of the ACM
Optimal Auctions Through Deep Learning
We overview recent research results that show how tools from deep learning are shaping up to become a powerful tool for the automated design of near-optimal auctions...
From Communications of the ACM
Technical Perspective: An Elegant Model for Deriving Equations
"Deriving Equations from Sensor Data Using Dimensional Function Synthesis," by Vasileios Tsoutsouras, et al., addresses the key problem of discovering relationships...
From Communications of the ACM
Deriving Equations from Sensor Data Using Dimensional Function Synthesis
We present a new method, which we call dimensional function synthesis, for deriving functions that model the relationship between multiple signals in a physical...
From Communications of the ACM
Technical Perspective: Race Logic Presents a Novel Form of Encoding
"In-Sensor Classification With Boosted Race Trees," by Georgios Tzimpragos, et al., proposes a surprising, novel, and creative approach to post-Moore's Law computing...
From Communications of the ACM
In-Sensor Classification With Boosted Race Trees
We demonstrate the potential of a novel form of encoding, race logic, in which information is represented as the delay in the arrival of a signal.
From Communications of the ACM
Technical Perspective: A Chiplet Prototype System for Deep Learning Inference
"Simba," by Yakun Sophia Shao, et al., presents a scalable deep learning accelerator architecture that tackles issues ranging from chip integration technology to...
From Communications of the ACM
Simba: Scaling Deep-Learning Inference with Chiplet-Based Architecture
This work investigates and quantifies the costs and benefits of using multi-chip-modules with fine-grained chiplets for deep learning inference, an application...
From Communications of the ACM
Technical Perspective: Why Don't Today's Deep Nets Overfit to Their Training Data?
"Understanding Deep Learning (Still) Requires Rethinking Generalization," Chiyuan Zhang, et al., brings a fundamental new theoretical challenge: Why don't today's...
From Communications of the ACM
Understanding Deep Learning (Still) Requires Rethinking Generalization
In this work, we presented a simple experimental framework for interrogating purported measures of generalization.
From Communications of the ACM
Technical Perspective: XNOR-Networks – Powerful but Tricky
How to produce a convolutional neural net that is small enough to run on a mobile device, and accurate enough to be worth using? The strategies in "Enabling AI...
From Communications of the ACM
Enabling AI at the Edge with XNOR-Networks
We present a novel approach to running state-of-the-art AI algorithms in edge devices, and propose two efficient approximations to standard convolutional neural...
From Communications of the ACM
Technical Perspective: When the Adversary Is Your Friend
The key insight of the "Generative Adversarial Networks," by Ian Goodfellow et al., is to learn a generative model's loss function at the same time as learning...
From Communications of the ACM
Generative Adversarial Networks
In this overview paper, we describe one particular approach to unsupervised learning via generative modeling called generative adversarial networks. We briefly...
From Communications of the ACM