We demonstrate longitudinal beam-steering with a 1×16 silicon optical phased array (OPA) using a monochromatic light source and thermo-optic control of the refractive index in the grating radiator region. The refractive index is controlled by forming a series of n-i-n heaters, placing i-regions in each radiator of the OPA. When the biased voltage in the heaters is increased, the refractive index of the radiator region is increased by the thermo-optic effect, and the longitudinal radiation angle is changed according to the Bragg condition. The transversal beam-steering is accomplished by phase control with the phase shifters, which are devised with a p-i-n diode using the electro-optic effect. With these electro-optic p-i-n phase shifters and n-i-n thermo-optic radiators, we achieve a relatively wide 2D beam-steering in a range of 10.0°/45.4° in the longitudinal/transversal directions with a 1.55 μm light source. The tuning efficiency is 0.016°/mW in the longitudinal beam-steering.
Networks-on-Chips (NoCs) provide communication platforms to Systems-on-Chips (SoCs). In NoCs, channels are generally shared between traffic flows, resulting in contention. However, certain flows require delivery guarantees. Differentiated quality-of-service (QoS) is achieved by providing guaranteed services such as guaranteed throughput (GT) to certain flows, on top of the regular best-effort (BE) delivery. Most current design methodologies employ a single-path mapping for each traffic flow. Such resource allocation is suboptimal for GT traffic and severely degrades performance of BE traffic. To solve this problem, we propose spatially distributing traffic and its guarantees in several paths per flow, while keeping the BE routing simple. Our experiments on a mesh in a transaction level simulator indicate that improvements of 20-48% in latency and 6-20% in throughput of BE traffic are achievable. We argue that this gain can be achieved cheaply with simple modifications to current methodologies.
Due to the black-box nature of deep networks, making explanations of their decision-making is extremely challenging. A solution is using post-hoc attention mechanisms with the deep network to verify the decision basis. However, those methods have problems such as gradient noise and false confidence. In addition, existing saliency methods either have limited performance by using only the last convolution layer or suffer from large computational overhead. In this work, we propose the Collection-CAM, which generates an attention map with low computational overhead while utilizing multi-level feature maps. First, the Collection-CAM searches for the most appropriate form of the partition through bottom-up clustering and clustering validation process. Then the Collection-CAM applies different pre-processing procedures on the shallow feature map and final feature map to overcome the false positiveness when applied without distinction. By combining collection-wise masks according to their contribution to the confidence score, the Collection-CAM completes the attention map generation process. Experimental results on ImageNet1k, UC Merced, and CUB dataset and various deep network models demonstrate that the Collection-CAM not only can synthesize a saliency map with a better visual explanation but also requires significantly lower computational overhead compared to those of region-based saliency methods.
Optical phased array (OPA) is considered as promising device in LiDAR application. We implemented a 1x16 silicon OPA consisting of an array of p-i-n electro-optic phase shifters and thermo-optic tunable grating radiators capable of two-dimensional beam-steering. The OPA was fabricated with CMOS-compatible process using SOI wafer. The p-i-n electro-optic phase shifters were formed in OPA channels for transversal beam-steering. With an array pitch of 2 μm, we attained transversal steering up to 45.6° at 1550 nm wavelength. For longitudinal beam-steering, we employed thermo-optic tunable grating radiators with p-i-n junction. The i-region covers whole radiator array and the p- and n-doped regions are placed on the both sides of the radiator array. This structure can provide fairly uniform heating of the radiator region, shifting the overall radiation field in longitudinal direction by the thermo-optic effect. As a result, a longitudinal beam-steering up to 10.3° was achieved by forward-biasing with a power consumption of 178 mW. This result proves a possibility of wide two-dimensional beam-steering with one-dimensional OPA without using tunable light source. We confirmed that the longitudinal tuning range obtained above is corresponding to near 100 nm wavelength tuning. Our device scheme can be a cost-effective solution of the OPA and also be a solution of self-adjustment for fluctuation of the wavelength-dependent performances.
We present a 1 × 16 silicon optical phased array using electro-optic phase shifters to attain high-speed operation with low power consumption. The phase shifters are constructed using a p-i-n junction structure with the pand n-doped regions formed on both sides of the i-region. The i-region width of the phase shifter is optimized considering the power consumption for phase-tuning and the propagation loss in the phase shifter. The fabricated p-i-n phase shifter exhibits a fast operating speed of 20 MHz and a low phase-tuning power of 1.7 mW/ π. From a 1-D optical phased array integrated with 2-μm-pitch grating radiators, a wide beam-steering range of 45° is attained along the transversal direction at a 1.55 μm wavelength. Average power consumption for the beam-forming operation of the 1 × 16 OPA is measured to be 39.6 mW and the average transition time to steer the beam is 24 ns.
Although several natural image datasets provide promising detection results, the performance of state-of-the-art detectors on aerial images is unsatisfactory in both accuracy and efficiency. To enhance the detection performance for aerial images, clustered object detection was introduced. It enhanced detection performance by identifying areas where objects are dense, which we call them as cluster chips, and applying fine detectors to those areas to combine them with global detection results. Although it demonstrated enhanced detection performance, still there is a room for improvement in terms of cluster chip selection. In this work, we propose a cluster chip selection scheme which identifies the cluster chip with greater performance improvement. We demonstrate our proposed method through comparison with other methods in terms of detection performance(e.g. mAP, mAP50 and mAP75). Experimental results show that the proposed method has performance gain over baseline methods in mAP, mAP50 and mAP75.
In a cloud environment, it is important for cloud broker to provide a cost-effective VM utilization. In this paper, we suggest a predicting scheme that can be applied for RVM provision by calculating demands. And there are some resource difference with respect to user’s needs on the process measuring clients’ needs. We also propose a method called M-C-VMA to handle the cost caused by the difference between real user demand and RVM provision. Performance evaluation showed that the proposed heuristic with VM Replacement is more efficient than C-VMA in cost performance. When M-C-VMA works on the VM allocation procedure, the result shows the higher RVM utilization than the not-modified method and consequently, it can lead the cost-efficient operation in broker system.
'%3e%3cg id='b'%3e%3cpath id='a' stroke='%23000' stroke-width='2' d='M-6-25H6m-12 3H6m-12 3H6'/%3e%3cuse xlink:href='%23a' y='44'/%3e%3c/g%3e%3cpath stroke='%23fff' d='M0 17v10'/%3e%3ccircle r='12' fill='%23cd2e3a'/%3e%3cpath fill='%230047a0' d='M0-12A6 6 0 0 0 0 0a6 6 0 0 1 0 12 12 12 0 0 1 0-24z'/%3e%3c/g%3e%3cg transform='rotate(-123.69)'%3e%3cuse xlink:href='%23b'/%3e%3cpath stroke='%23fff' d='M0-23.5v3M0 17v3.5m0 3v3'/%3e%3c/g%3e%3c/svg%3e)
