1k citations · 1.2k across the 11 of their papers we have counts for
13 papers
Ps and Qs: Quantization-aware pruning for efficient low latency neural network inference
Benjamin Hawks, Javier Duarte, Nicholas J. Fraser +3
Efficient machine learning implementations optimized for inference in hardware have wide-ranging benefits, depending on the application, from lower inference latency to higher data…
Memory-Efficient Dataflow Inference for Deep CNNs on FPGA
Lucian Petrica, Tobias Alonso, Mairin Kroes +3
Custom dataflow Convolutional Neural Network (CNN) inference accelerators on FPGA are tailored to a specific CNN topology and store parameters in On-Chip Memory (OCM), resulting in…
FAT: Training Neural Networks for Reliable Inference Under Hardware Faults
Ussama Zahid, Giulio Gambardella, Nicholas J. Fraser +2
Deep neural networks (DNNs) are state-of-the-art algorithms for multiple applications, spanning from image classification to speech recognition. While providing excellent accuracy,…
LogicNets: Co-Designed Neural Networks and Circuits for Extreme-Throughput Applications
Yaman Umuroglu, Yash Akhauri, Nicholas J. Fraser +1
Deployment of deep neural networks for applications that require very high throughput or extremely low latency is a severe computational challenge, further exacerbated by inefficie…
FINN-R: An End-to-End Deep-Learning Framework for Fast Exploration of Quantized Neural Networks
Michaela Blott, Thomas Preusser, Nicholas Fraser +3
Convolutional Neural Networks have rapidly become the most successful machine learning algorithm, enabling ubiquitous machine vision and intelligent decisions on even embedded comp…
Accuracy to Throughput Trade-offs for Reduced Precision Neural Networks on Reconfigurable Logic
Jiang Su, Nicholas J. Fraser, Giulio Gambardella +5
Modern CNN are typically based on floating point linear algebra based implementations. Recently, reduced precision NN have been gaining popularity as they require significantly les…
SYQ: Learning Symmetric Quantization For Efficient Deep Neural Networks
Julian Faraone, Nicholas Fraser, Michaela Blott +1
Inference for state-of-the-art deep neural networks is computationally expensive, making them difficult to deploy on constrained hardware environments. An efficient way to reduce t…
Scaling Neural Network Performance through Customized Hardware Architectures on Reconfigurable Logic
Michaela Blott, Thomas B. Preusser, Nicholas Fraser +4
Convolutional Neural Networks have dramatically improved in recent years, surpassing human accuracy on certain problems and performance exceeding that of traditional computer visio…
Inference of Quantized Neural Networks on Heterogeneous All-Programmable Devices
Thomas B. Preußer, Giulio Gambardella, Nicholas Fraser +1
Neural networks have established as a generic and powerful means to approach challenging problems such as image classification, object detection or decision making. Their successfu…
Quantizing Convolutional Neural Networks for Low-Power High-Throughput Inference Engines
Sean O. Settle, Manasa Bollavaram, Paolo D'Alberto +6
Deep learning as a means to inferencing has proliferated thanks to its versatility and ability to approach or exceed human-level accuracy. These computational models have seemingly…
Compressing Low Precision Deep Neural Networks Using Sparsity-Induced Regularization in Ternary Networks
Julian Faraone, Nicholas Fraser, Giulio Gambardella +2
A low precision deep neural network training technique for producing sparse, ternary neural networks is presented. The technique incorporates hard- ware implementation costs during…
Scaling Binarized Neural Networks on Reconfigurable Logic
Nicholas J. Fraser, Yaman Umuroglu, Giulio Gambardella +4
Binarized neural networks (BNNs) are gaining interest in the deep learning community due to their significantly lower computational and memory cost. They are particularly well suit…
FINN: A Framework for Fast, Scalable Binarized Neural Network Inference
Yaman Umuroglu, Nicholas J. Fraser, Giulio Gambardella +4
Research has shown that convolutional neural networks contain significant redundancy, and high classification accuracy can be obtained even when weights and activations are reduced…