Convolutional neural network (CNN) delivers impressive achievements in computer vision and machine learning field. However, CNN incurs high computational complexity, especially for vision quality applications because of large image resolution. In this paper, we propose an iterative architecture-aware pruning algorithm with adaptive magnitude threshold while cooperating with quality-metric measurement simultaneously. We show the performance improvement applied on vision quality applications and provide comprehensive analysis with flexible pruning configuration. With the proposed method, the Multiply-Accumulate (MAC) of state-of-the-art low-light imaging (SID) and super-resolution (EDSR) are reduced by 58% and 37% without quality drop, respectively. The memory bandwidth (BW) requirements of convolutional layer can be also reduced by 20% to 40%.

Majority of tasks are medium and light loads

In this paper, a microcell is designed for enhancing the uplink capacity of an existing terrestrial cellular code division multiple access (CDMA) system. The proposed designs involve integration of a high altitude platform station (HAPS) and the terrestrial system. By adjusting the projection of a phased array antenna onboard the HAPS, two designs of microcell are developed in the integrated system. One is that main lobe directions of the antenna coincide with the locations of terrestrial base stations; the other instead has a projection pattern with main lobes pointing toward terrestrial cell boundaries. Numerical analysis is performed and the results show that the uplink capacity is significantly enhanced under both our proposed designs, compared with a purely ground-based or a purely HAPS CDMA system; furthermore, the latter design can provide more than 15% higher uplink capacity over the former one.

To reduce area of texture filter for real-time rendering is challenging as user demand for 3-D scene realism increases. This work presents a low-area run-time reconfigurable texture filter design given certain performance requirement. A texture filter is a part of a texture unit. The reconfiguration is on a number of bilinear filters as basic function blocks, can be implemented within individual texture unit or across multiple texture units, and is based on hierarchical decomposition of all kinds of filtering algorithms. Two reconfiguration designs are proposed and examined: The self-content design, in which a texture unit must be equipped with all necessary hardware to deal with all different filtering algorithms; and the sharable design, in which texture units can share their texture filters (and the total amount of useable texture units may thus change). Experimental results show that the sharable design achieves a 12% area reduction, as compared with the self-content design.

In a cellular code division multiple access (CDMA) system, uplink capacity is limited by interference. For a ground-based CDMA system or a high altitude platform station (HAPS) CDMA system, the interference is contributed by users in the same cell and in other cells. In this paper, a novel scheme is proposed to reduce both the same-cell and other-cell interferences in a cellular CDMA system. The proposed scheme attempts to lower the required transmitting power for each user by integrating terrestrial base stations (BSs) with a HAPS to provide two separate paths for signal reception. Numerical results show that the total interference is significantly reduced and the uplink capacity is improved under the proposed scheme, compared to either a HAPS CDMA system or a ground-based CDMA system.

High altitude platform stations (HAPSs) are widely known as an alternative means to provide broadband applications and code division multiple access (CDMA) based international mobile telecommunications-2000 (IMT-2000) communications. Previous studies have shown that platform instability caused by stratospheric winds may lead to system level performance degradation. In this paper, we propose to combat this problem by using multiple HAPSs to provide diversity of signal reception. Though the proposed scheme would enhance the reliability of service provision, the required uplink transmitting power may need to be increased to ensure that every serving HAPS receives at least the target receive power. This will elevate the interference level and may adversely affect the uplink capacity. Numerical analysis is thus performed to investigate the impact of using multiple HAPSs on the uplink capacity.

In this paper, we consider the uplink capacity of an integrated high altitude platform station (HAPS)-terrestrial code division multiple access (CDMA) system in which sharing band overlay configuration is used. To increase system capacity, an enhancement is investigated by assuming that directional antenna is used for high mobility users to access to the HAPS. This will lead to a reduction of not only the interference from HAPS to terrestrial, but also the interference from terrestrial to HAPS as the uplink power to HAPS can accordingly be increased. Numerical results are obtained analytically and it is shown that the system capacity can be enhanced significantly under the proposed scheme

In this paper, a general reliability-updating formula is proposed for trellis-based soft-output decoding algorithms to optimize the tradeoff between performance and complexity. Based on the general formula, new algorithms are presented, and a concise interpretation is provided to relate new proposed algorithms and those reported previously. In addition, we devise a unified decoding structure with respect to the general formula. All trellis-based algorithms are mapped into a single decoding process under the unified structure, and different algorithms can be easily switched between one another. Owing to the modularity and flexibility, this unified structure is especially suitable for turbo decoders with programmable implementation.