Abstract-Convolutional neural network (CNN) offers significant accuracy in image detection. To implement imagedetection using CNN in the internet of things (IoT) devices, a streaming hardware accelerator is proposed. The proposed accelerator optimizes the energy efficiency by avoiding unnecessary data movement. With unique filter decomposition technique, the accelerator can support arbitrary convolution window size. In addition, max pooling function can be computed in parallel with convolution by using separate pooling unit, thus achieving throughput improvement. A prototype accelerator was implemented in TSMC 65nm technology with a core size of 5mm 2 . The accelerator can support major CNNs and achieve 152GOPS peak throughput and 434GOPS/W energy efficiency at 350mW, making it a promising hardware accelerator for intelligent IoT devices.

For optoelectronic applications, colloidal CdSe quantum dots (QDs) have been integrated into solid devices by using optically transparent polymer matrices that embedded the colloidal QDs. We systematically studied the effect of annealing and photoactivation on the band-band (BB) and surface trap state (STS) transitions of colloidal CdSe QDs embedded in polymethylmethacrylate (QDs-PMMA). The QDs-PMMA composites demonstrate enhancement of the STS emissions while their annealing leads to an intensity quench of both BB and STS emissions. The annealing process also causes the red shift of the BB emission. By contrast, photoactivation of QDs-PMMA composites results in the remarkable recovery of luminescence intensity accompanied by blue shift of the emissions. Furthermore, it is found in the photoactivation process that the STS emission can be saturated earlier than the BB emission, which renders it possible to tune the light color of the emissions. The combination of annealing and photoactivation could undoubtedly provide an effective way to precisely tune the colors of light-emitting devices that use colloidal CdSe QDs.

We explore a strongly interacting QDs/Ag plasmonic coupling structure that enables multiple approaches to manipulate light emission from QDs. Group II-VI semiconductor QDs with unique surface states (SSs) impressively modify the plasmonic character of the contiguous Ag nanostructures whereby the localized plasmons (LPs) in the Ag nanostructures can effectively extract the non-radiative SSs of the QDs to radiatively emit via SS-LP resonance. The SS-LP coupling is demonstrated to be readily tunable through surface-state engineering both during QD synthesis and in the post-synthesis stage. The combination of surface-state engineering and band-tailoring engineering allows us to precisely control the luminescence color of the QDs and enables the realization of white-light emission with single-size QDs. Being a versatile metal, the Ag in our optical device functions in multiple ways: as a support for the LPs, for optical reflection, and for electrical conduction. Two application examples of the QDs/Ag plasmon coupler for optical devices are given, an Ag microcavity + plasmon-coupling structure and a new QD light-emitting diode. The new QDs/Ag plasmon coupler opens exciting possibilities in developing novel light sources and biomarker detectors.

1  Abstract-An analog neural network computing engine based on CMOS-compatible charge-trap transistor (CTT) is proposed in this paper. CTT devices are used as analog multipliers. Compared to digital multipliers, CTT-based analog multiplier shows significant area and power reduction. The proposed computing engine is composed of a scalable CTT multiplier array and energy efficient analog-digital interfaces. By implementing the sequential analog fabric (SAF), the engine's mixed-signal interfaces are simplified and hardware overhead remains constant regardless of the size of the array. A proof-of-concept 784 by 784 CTT computing engine is implemented using TSMC 28nm CMOS technology and occupies 0.68mm 2 . The simulated performance achieves 76.8 TOPS (8-bit) with 500 MHz clock frequency and consumes 14.8 mW. As an example, we utilize this computing engine to address a classic pattern recognition problem -classifying handwritten digits on MNIST database and obtained a performance comparable to state-of-the-art fully connected neural networks using 8-bit fixed-point resolution. Index Terms-Artificial neural networks, Charge-trap transistors, Fully-connected neural networks, Analog computing engine Recently, charge-trap transistors (CTTs) were demonstrated to be used as digital memory devices in [19-20] with

No abstract

No abstract

No abstract

No abstract