Autonomous driving services depends on active sensing from modules such as camera,
LiDAR, radar, and communication units. Traditionally, these modules process the sensed …

Embedded inference engines for convolutional networks must be parsimonious in memory
bandwidth and buffer sizing to meet power and cost constraints. We present an analytical …

This paper introduces the NVIDIA deep learning accelerator (NVDLA), including its
hardware architecture specification and software environment. At the same time, the basic …

We present a 3.1 POp/s/W fully digital hardware accelerator for ternary neural networks
(TNNs). CUTIE, the completely unrolled ternary inference engine, focuses on minimizing …

Always-on keyword spotting (KWS) hardware is gaining popularity in ultra-low power IoT
applications where specific words are used to wake up and activate the power hungry …

The superiority of various Deep Neural Networks (DNN) models, such as Convolutional
Neural Networks (CNN), Generative Adversarial Networks (GAN), and Recurrent Neural …

Neural network pruning allows for impressive theoretical reduction of models sizes and
complexity. However it usually offers little practical benefits as it is most often limited to just …

We introduce Amortized Neural Networks (AmNets), a compute cost-and latency-aware
network architecture particularly well-suited for sequence modeling tasks. We apply AmNets …

Keyword Spotting (KWS) has been the subject of research in recent years given the increase
of embedded systems for command recognition such as Alexa, Google Home, and Siri …

Deep neural networks are getting larger and, therefore, harder to deploy on constrained IoT
devices. Split computing provides a solution by splitting a network and placing the first few …