Exploiting Symmetric Temporally Sparse BPTT for Efficient RNN Training

Recurrent Neural Networks (RNNs) are useful in temporal sequence tasks. However, training RNNs involves dense matrix multiplications which require hardware that can support a large number of arithmetic operations and memory accesses. Implementing online training of RNNs on the edge calls for optimized algorithms for an efficient deployment on...

Structure and dynamics of urban freight truck movements: A complex network theory perspective

Knowledge of the core structure and inherent dynamics of urban freight transport systems is important for the development of policies, aimed at improving the livability and sustainability of cities. The past decade has witnessed a great deal of efforts into analyzing the geographic structure of urban freight transport systems. However, in...

An Area-Efficient Ultra-Low-Power Time-Domain Feature Extractor for Edge Keyword Spotting

Keyword spotting (KWS) is an important task on edge low-power audio devices. A typical edge KWS system consists of a front-end feature extractor which outputs mel-scale frequency cepstral coefficients (MFCC) features followed by a back-end neural network classifier. KWS edge designs aim for the best power-performance-area metrics. This work...

Analysis of the cohesion/adhesion proportion around bitumen-mineral failure interface under tensile loading

The occurrence of cohesive/adhesive hybrid failure at the bitumen-aggregate interface is widely acknowledged, however, the quantitative evaluation of the cohesion/adhesion proportion is relatively weak. This study explored cohesive/adhesive hybrid failure and provided a quantitative analysis of the proportion between cohesion and adhesion....

Nanoprobes for PET/MR Imaging

The development of clinical imaging techniques significantly improves diagnostic accuracy and provides guidance for personalized treatment of individuals. However, every single imaging modality has its distinct drawbacks that cannot fully fulfil the diagnosis requirement. Thus, rational combination of different imaging modalities can achieve...

3ET: Efficient Event-based Eye Tracking using a Change-Based ConvLSTM Network

This paper presents a sparse Change-Based Convolutional Long Short-Term Memory (CB-ConvLSTM) model for event-based eye tracking, key for next-generation wearable healthcare technology such as AR/VR headsets. We leverage the benefits of retina-inspired event cameras, namely their low-latency response and sparse output event stream, over...

Research on Design and Performance of Self-Compacting Cement Emulsified Bitumen Mixture (CEBM)

To meet the needs of the road industry for maintenance operations, a new cement emulsified bitumen mixture (CEBM) with early-strength, self-compacting, and room-temperature construction characteristics was designed. The strength formation mechanism of CEBM was revealed with a scanning electron microscope (SEM) and the surface free energy (SFE...

Aqueous chemoenzymatic one-pot enantioselective synthesis of tertiary α-aryl cycloketonesviaPd-catalyzed C-C formation and enzymatic C=C asymmetric hydrogenation

An aqueous chemoenzymatic cascade reaction combining Pd-catalyzed C-C formation and enzymatic C=C asymmetric hydrogenation (AH) was developed for enantioselective synthesis of tertiary α-aryl cycloketones in good yields and excellent enantioselectivities. The stereopreference of the enzyme in AH of α-aryl cyclohexenones was studied. An...

New innovations in pavement materials and engineering: A review on pavement engineering research 2021

Sustainable and resilient pavement infrastructure is critical for current economic and environmental challenges. In the past 10 years, the pavement infrastructure strongly supports the rapid development of the global social economy. New theories, new methods, new technologies and new materials related to pavement engineering are emerging....