Browsing by Author "Wu, Wei-Liang"

Now showing 1 - 4 of 4

Corner-Stitching-Based Multilayer Obstacle-Avoiding Component-to-Component Rectilinear Minimum Spanning Tree Construction
(2020-03-01)
Su, Yen-Yu
;
Wang, Shuo-Hui
;
Wu, Wei-Liang
;
Lin, Mark Po-Hung
;
電子工程學系及電子研究所
Engineering change orders (ECOs) are pervasively applied to modern physical design of nanometer integrated circuits for cost-effective design changes. After applying ECO, a net may become open, which results in a large number of disconnected net components. Each net component further consists of a set of connected net shapes and vias on different layers. It is very challenging to efficiently and effectively identify all disconnected net components and find an obstacle-avoiding minimal-cost routing path among those net components. This paper introduces an open-net finder and an open-net connector based on the corner-stitching data structure for open-net routing, and proposes a new method of constructing a multilayer obstacle-avoiding component-to-component rectilinear minimum spanning tree. The preliminary idea and implementation of the proposed method had received the first place award in ICCAD-2017 CAD contest. This paper further details the complete idea of the proposed method. Experimental results show that the proposed open-net finder and connector outperform the top three open-net routers in the ICCAD-2017 CAD contest and two latest published works, with better routing costs and much shorter runtime.
Locality-constrained max-margin sparse coding
(2017-05-01)
Hsaio, Wen-Hoar
;
Liu, Chien-Liang
;
Wu, Wei-Liang
;
資訊工程學系
;
工業工程與管理學系
This work devises a locality-constrained max-margin sparse coding (LC-MMSC) framework, which jointly considers reconstruction loss and hinge loss simultaneously. Traditional sparse coding algorithms use l(1) constraint to force the representation to be sparse, leading to computational expensive process to optimize the objective function. This work uses locality constraint in the framework to preserve information of data locality and avoid the optimization of l(1). The obtained representation can achieve the goal of data locality and sparsity. Additionally, this work optimizes coefficients, dictionaries and classification parameters simultaneously, and uses block coordinate descent to learn all the components of the proposed model. This work uses semi supervised learning approach in the proposed framework, and the goal is to use both labeled data and unlabeled data to achieve accurate classification performance and improve the generalization of the model. We provide theoretical analysis on the convergence of the proposed LC-MMSC algorithm based on Zangwill's global convergence theorem. This work conducts experiments on three real datasets, including Extended YaleB dataset, AR face dataset and Caltech101 dataset. The experimental results indicate that the proposed algorithm outperforms other comparison algorithms.
Maximum-margin sparse coding
(2017-05-17)
Liu, Chien-Liang
;
Hsaio, Wen-Hoar
;
Xiao, Bin
;
Chen, Chun-Yu
;
Wu, Wei-Liang
This work devises a maximum-margin sparse coding algorithm, jointly considering reconstruction loss and hinge loss in the model. The sparse representation along with maximum-margin constraint is analogous to kernel trick and maximum-margin properties of support vector machine (SVM), giving a base for the proposed algorithm to perform well in classification tasks. The key idea behind the proposed method is to use labeled and unlabeled data to learn discriminative representations and model parameters simultaneously, making it easier to classify data in the new space. We propose to use block coordinate descent to learn all the components of the proposed model and give detailed derivation for the update rules of the model variables. Theoretical analysis on the convergence of the proposed MMSC algorithm is provided based on Zangwill's global convergence theorem. Additionally, most previous research studies on dictionary learning suggest to use an overcomplete dictionary to improve classification performance, but it is computationally intensive when the dimension of the input data is huge. We conduct experiments on several real data sets, including Extended YaleB, AR face, and Caltech101 data sets. The experimental results indicate that the proposed algorithm outperforms other comparison algorithms without an over-complete dictionary, providing flexibility to deal with high-dimensional data sets. (C)2017 Elsevier B.V. All rights reserved.
Using pre-distorted PAM-4 signal and parallel resistance circuit to enhance the passive solar cell based visible light communication
(2018-01-15)
Wang, Hao-Yu
;
Wu, Jhao-Ting
;
Chow, Chi-Wai
;
Liu, Yang
;
Yeh, Chien-Hung
Using solar cell (or photovoltaic cell) for visible light communication (VLC) is attractive. Apart from acting as a VLC receiver (Rx), the solar cell can provide energy harvesting. This can be used in self-powered smart devices, particularly in the emerging "Internet of Things (IoT)'' networks. Here, we propose and demonstrate for the first time using pre-distortion pulse-amplitude-modulation (PAM)-4 signal and parallel resistance circuit to enhance the transmission performance of solar cell Rx based VLC. Pre-distortion is a simple non-adaptive equalization technique that can significantly mitigate the slow charging and discharging of the solar cell. The equivalent circuit model of the solar cell and the operation of using parallel resistance to increase the bandwidth of the solar cell are discussed. By using the proposed schemes, the experimental results show that the data rate of the solar cell Rx based VLC can increase from 20 kbit/s to 1.25 Mbit/s (about 60 times) with the bit error-rate (BER) satisfying the 7% forward error correction (FEC) limit. (C) 2017 Elsevier B.V. All rights reserved.