Efficiency analysis for stochastic dynamic facility layout problem using meta-heuristic,data envelopment analysis and machine learning

The facility layout problem (FLP) is a combinatorial optimization problem. The performance of the layout design is significantly impacted by diverse, multiple factors. The use of algorithmic or procedural design methodology in ranking and identification of efficient layout is ineffective. In this context, this study proposes a three-stage methodology where data envelopment analysis (DEA) is augmented with unsupervised and supervised machine learning (ML). In stage 1, unsupervised ML is used for the clustering of the criteria in which the layouts need to be evaluated using homogeneity. Layouts are generated using simulated annealing, chaotic simulated annealing, and hybrid firefly algorithm/chaotic simulated annealing meta-heuristics. In stage 2, the nonparametric DEA approach is used to identify efficient and inefficient layouts. Finally, supervised ML utilizes the performance frontiers from DEA (efficiency scores) to generate a trained model for getting the unique rankings and predicted efficiency scores of layouts. The proposed methodology overcomes the limitations associated with large datasets that contain many inputs / outputs from the conventional DEA and improves the prediction accuracy of layouts. A Gaussian distribution product demand dataset for time period T = 5 and facility size N = 12 is used to prove the effectiveness of the methodology.     

Generative adversarial network for early-stage design flexibility in topology optimization for additive manufacturing

Topology optimization has become a valuable design tool for structures to be fabricated by additive manufacturing (AM). However, during early stage design, parameters are frequently evolving, resulting in multiple similar TO runs. Especially when design for manufacturing principles expand the parameter space, this iterative process is computationally burdensome, and does not take advantage of redundant information in each study. We introduce a deep learning-based framework that learns latent similarities between runs in a training set to predict near optimal designs, enabling efficient wholistic understanding of the problem setup space, which includes both loading conditions and, for the first time in this study, manufacturing process configuration. Learning was achieved using a conditional generative adversarial network (cGAN) trained on a dataset of randomized boundary conditions, loadings, and AM build orientations, and the corresponding optimal structures obtained through overhang-filtered TO. cGAN predictions showed good agreement with true optima. For even greater accuracy, predictions can be post-processed by applying a small number of TO iterations. Manifold learning techniques were used to provide further insight, and we were able to conclude that the cGAN error generally increases with distance between the load and the boundary conditions or build plate. Interestingly, in 9% of test cases, the cGAN generated structures with compliances better than the corresponding TO-calculated structures, often by as much as 50 % with an average of 7.8 %. That some of these structures appeared qualitatively different in form suggests the potential value of the approach in other domains such as generative design, where a range of alternate near-optimal designs are used to guide the ideation process.     

Banner layout retargeting with hierarchical reinforcement learning and variational autoencoder

In many advertising areas, banners are often generated with different display sizes, so designers have to make huge efforts to retarget their designs to each size. Automating such retargeting process can greatly save time for designers and let them put creativity on new ads. This paper proposes a hierarchical reinforcement learning-based (HRL-based) method and a variational autoencoder-based (VAE-based) method by treating the automated banner retargeting problem as a layout retargeting task. The HRL and VAE models are trained separately to learn the scaling and positioning policy of the design elements from an original (base) layout. Hence, the proposed method can generate appropriate layouts for different target banner sizes. Meanwhile, evaluation metrics are proposed to assess the quality of generated layouts and are also reward conditions during the training process. To evaluate performances of the two models, SOTA methods such as Non-linear Inverse Optimization (NIO), Triangle Interpolation (TI), and Layout GAN (LGAN) are implemented and compared. Experimental results show that both HRL- and VAE-based methods retarget design layouts effectively, and the VAE model achieves better performance than the HRL model.

     

A Comparison of User-Generated and Automatic Graph Layouts

The research presented in this paper compares user-generated and automatic graph layouts. Following the methods suggested by van Ham et al. (2008), a group of users generated graph layouts using both multi-touch interaction on a tabletop display and mouse interaction on a desktop computer. Users were asked to optimize their layout for aesthetics and analytical tasks with a social network. We discuss characteristics of the user-generated layouts and interaction methods employed by users in this process. We then report on a web-based study to compare these layouts with the output of popular automatic layout algorithms. Our results demonstrate that the best of the user-generated layouts performed as well as or better than the physics-based layout. Orthogonal and circular automatic layouts were found to be considerably less effective than either the physics-based layout or the best of the user-generated layouts. We highlight several attributes of the various layouts that led to high accuracy and improved task completion time, as well as aspects in which traditional automatic layout methods were unsuccessful for our tasks.     

基于CGAN的中国山水画布局可调的仿真生成方法

以往的山水画计算机仿真由于未从山水画整体布局的角度进行研究,难以实现完整的画作生成.针对上述问题,文中提出布局引导、可实现完整画作生成的中国山水画仿真方法.基于山水画的绘制特点设计可行的布局标签图结构,用于表达山水画的构图形态和要素.借鉴条件生成对抗网络(CGAN)的思想,针对山水画的布局和笔触特点,设计并训练多尺度特征融合的网络结构(MSFF-CGAN),实现布局标签图到仿真山水画这一异质生成过程.同时针对网络训练过程中布局标签图数据稀缺的问题,采用语义关联的颜色像素聚类算法快速生成标签图.为了提高生成图的艺术真实感,引入MemNet超分辨网络增强生成图的纹理细节.实验表明,文中方法生成的仿真山水画具有较好的完整性和艺术真实感,不仅可以应对简单的手绘涂鸦式草图,还可以通过在布局空间的编辑操作,达到对画作空间进行编辑的效果.     

基于自注意力机制的条件生成对抗网络

近年来,越来越多的生成对抗网络出现在深度学习的各个领域中.条件生成对抗网络(Conditional Generative Adver-sarial Networks,cGAN)开创性地将监督学习引入到无监督的GAN网络中,这使得GAN可以生成有标签数据.传统的GAN通过多次卷积运算来模拟不同区域之间的相关性,进而生成图...     

数据失衡下基于WGAN和GAPCNN的轴承故障诊断研究

针对轴承故障数据严重失衡导致所训练的模型诊断能力和泛化能力较差等问题,提出基于Wasserstein距离的生成对抗网络来平衡数据集的方法。该方法首先将少量故障样本进行对抗训练,待网络达到纳什均衡时,再将生成的故障样本添加到原始少量故障样本中起到平衡数据集的作用;提出基于全局平均池化卷积神经网络的诊断模型,将平衡后的数据集输入到诊断模型中进行训练,通过模型自适应地逐层提取特征,实现故障的精确分类诊断。实验结果表明,所提诊断方法优于其他算法和模型,同时拥有较强的泛化能力和鲁棒性。     

SVM based layout retargeting for fast and regularized inverse lithography

Inverse lithography technology （ILT）, also known as pixel-based optical proximity correction （PB-OPC）, has shown promising capability in pushing the current 193 nm lithography to its limit. By treating the mask optimization process as an inverse problem in lithography, ILT provides a more complete exploration of the solution space and better pattern fidelity than the tradi-tional edge-based OPC. However, the existing methods of ILT are extremely time-consuming due to the slow convergence of the optimization process. To address this issue, in this paper we propose a support vector machine （SVM） based layout retargeting method for ILT, which is designed to generate a good initial input mask for the optimization process and promote the convergence speed. Supervised by optimized masks of training layouts generated by conventional ILT, SVM models are learned and used to predict the initial pixel values in the‘undefined areas’ of the new layout. By this process, an initial input mask close to the final optimized mask of the new layout is generated, which reduces iterations needed in the following optimization process. Manu-facturability is another critical issue in ILT;however, the mask generated by our layout retargeting method is quite irregular due to the prediction inaccuracy of the SVM models. To compensate for this drawback, a spatial filter is employed to regularize the retargeted mask for complexity reduction. We implemented our layout retargeting method with a regularized level-set based ILT （LSB-ILT） algorithm under partially coherent illumination conditions. Experimental results show that with an initial input mask generated by our layout retargeting method, the number of iterations needed in the optimization process and runtime of the whole process in ILT are reduced by 70.8%and 69.0%, respectively.     

基于条件生成对抗网络的乳腺上皮和间质区域自动分割

乳腺病理组织图像中上皮和间质区域的自动分割对乳腺癌的诊断和治疗具有非常重要的临床意义。但是由于乳腺组织病理图像中上皮和间质区域具有高度复杂性,因此一般的分割模型很难只根据提供的分割标记来有效地训练,并对两种区域进行快速、准确的分割。为此,提出一种基于条件对抗网络（cGAN）的上皮和间质分割条件对抗网络（EPScGAN）模型。在EPScGAN中,判别器的判别机制为生成器的训练提供了一个可训练的损失函数,来更加准确地衡量出生成器网络的分割结果输出和真实标记之间的误差,从而更好地指导生成器的训练。从荷兰癌症研究所（NKI）和温哥华综合医院（VGH）两个机构提供的专家标记的乳腺病理图像数据集中随机裁剪出1 286张尺寸为512×512的图像作为实验数据集,然后将该数据集按照7：3的比例划分为训练集和测试集对EPScGAN模型进行训练和测试。结果表明,EPScGAN模型在测试集的平均交并比（mIoU）为78.12%,和其他6种流行的深度学习分割模型相比较,提出的EPScGAN具有更好的分割性能。     

Visualization and User-Modeling for Browsing Personal Photo Libraries

We present a user-centric system for visualization and layout for content-based image retrieval. Image features (visual and/or semantic) are used to display retrievals as thumbnails in a 2-D spatial layout or “configuration” which conveys all pair-wise mutual similarities. A graphical optimization technique is used to provide maximally uncluttered and informative layouts. Moreover, a novel subspace feature weighting technique can be used to modify 2-D layouts in a variety of context-dependent ways. An efficient computational technique for subspace weighting and re-estimation leads to a simple user-modeling framework whereby the system can learn to display query results based on layout examples (or relevance feedback) provided by the user. The resulting retrieval, browsing and visualization can adapt to the user's (time-varying) notions of content, context and preferences in style and interactive navigation. Monte Carlo simulations with machine-generated layouts as well as pilot user studies have demonstrated the ability of this framework to model or “mimic” users, by automatically generating layouts according to their preferences.     

Conditional Generative Adversarial Network Approach for Autism Prediction

Autism Spectrum Disorder (ASD) requires a precise diagnosis in order to be managed and rehabilitated. Non-invasive neuroimaging methods are disease markers that can be used to help diagnose ASD. The majority of available techniques in the literature use functional magnetic resonance imaging (fMRI) to detect ASD with a small dataset, resulting in high accuracy but low generality. Traditional supervised machine learning classification algorithms such as support vector machines function well with unstructured and semi structured data such as text, images, and videos, but their performance and robustness are restricted by the size of the accompanying training data. Deep learning on the other hand creates an artificial neural network that can learn and make intelligent judgments on its own by layering algorithms. It takes use of plentiful low-cost computing and many approaches are focused with very big datasets that are concerned with creating far larger and more sophisticated neural networks. Generative modelling, also known as Generative Adversarial Networks (GANs), is an unsupervised deep learning task that entails automatically discovering and learning regularities or patterns in input data in order for the model to generate or output new examples that could have been drawn from the original dataset. GANs are an exciting and rapidly changing field that delivers on the promise of generative models in terms of their ability to generate realistic examples across a range of problem domains, most notably in image-to-image translation tasks and hasn't been explored much for Autism spectrum disorder prediction in the past. In this paper, we present a novel conditional generative adversarial network, or cGAN for short, which is a form of GAN that uses a generator model to conditionally generate images. In terms of prediction and accuracy, they outperform the standard GAN. The proposed model is 74% more accurate than the traditional methods and takes only around 10 min for training even with a huge dataset.     

Perfect Matching Quad Layouts for Manifold Meshes

This paper introduces a new approach to automatically generate pure quadrilateral patch layouts on manifold meshes. The algorithm is based on a careful construction of a singularity graph of a given input frame field or a given periodic global parameterization. A pure quadrilateral patch layout is then derived as a constrained minimum weight perfect matching of that graph. The resulting layout is optimal relative to a balance between coarseness and geometric feature alignment. We formulate the problem of finding pure quadrilateral patch layouts as a global optimization problem related to a well‐known concept in graph theory. The main advantage of the new method is its simplicity and its computation speed. Patch layouts generated by the present algorithm are high quality and are very competitive compared to current state of the art.     

基于条件生成式对抗网络的数据增强方法

深度卷积神经网络（CNN）在大规模带有标签的数据集训练下,训练后模型能够取得高的识别率或好的分类效果,而利用较小规模数据集训练CNN模型则通常出现过拟合现象。针对这一问题,提出了一种集成高斯混合模型（GMM）及条件生成式对抗网络（CGAN）的数据增强方法并记作GMM-CGAN。首先,通过围绕核心区域随机滑动采样的方法增加数据集样本数量;其次,假定噪声随机向量服从GMM描述的分布,将它作为CGAN生成器的初始输入,图像标签作为CGAN条件,训练CGAN以及GMM模型的参数;最后,利用已训练CGAN生成符合样本真实分布的新数据集。对包含12种雾型386个样本的天气形势图基准集利用GMM-CGAN方法进行数据增强,增强后的数据集样本数多达38600个,将该数据集训练的CNN模型与仅使用仿射变换增强的数据集及CGAN方法增强的数据集训练的CNN模型相比,实验结果表明,前者的平均分类正确率相较于后两个模型分别提高了18.2%及14.1%,达到89.1%。     

高分辨卫星图像卷积神经网络分类模型

目的 卫星图像往往目标、背景复杂而且带有噪声,因此使用人工选取的特征进行卫星图像的分类就变得十分困难。提出一种新的使用卷积神经网络进行卫星图像分类的方案。使用卷积神经网络可以提取卫星图像的高层特征,进而提高卫星图像分类的识别率。方法 首先,提出一个包含六类图像的新的卫星图像数据集来解决卷积神经网络的有标签训练样本不足的问题。其次,使用了一种直接训练卷积神经网络模型和3种预训练卷积神经网络模型来进行卫星图像分类。直接训练模型直接在文章提出的数据集上进行训练,预训练模型先在ILSVRC（the ImageNet large scale visual recognition challenge）-2012数据集上进行预训练,然后在提出的卫星图像数据集上进行微调训练。完成微调的模型用于卫星图像分类。结果 提出的微调预训练卷积神经网络深层模型具有最高的分类正确率。在提出的数据集上,深层卷积神经网络模型达到了99.50%的识别率。在数据集UC Merced Land Use上,深层卷积神经网络模型达到了96.44%的识别率。结论 本文提出的数据集具有一般性和代表性,使用的深层卷积神经网络模型具有很强的特征提取能力和分类能力,且是一种端到端的分类模型,不需要堆叠其他模型或分类器。在高分辨卫星图像的分类上,本文模型和对比模型相比取得了更有说服力的结果。     

A Novel Two-stage Learning Pipeline for Deep Neural Networks

In this work, a training method was proposed for Deep Neural Networks (DNNs) based on a two-stage structure. Local DNN models are trained in all local machines and uploaded to the center with partial training data. These local models are integrated as a new DNN model (combination DNN). With another DNN model (optimization DNN) connected, the combination DNN forms a global DNN model in the center. This results in greater accuracy than local DNN models with smaller amounts of data uploaded. In this case, the bandwidth of the uploaded data is saved, and the accuracy is maintained as well. Experiments are conducted on MNIST dataset, CIFAR-10 dataset and LFW dataset. The results show that with less training data uploaded, the global model produces greater accuracy than local models. Specifically, this method focuses on condition of big data.     

Big Data Analytics: Deep Content-Based Prediction with Sampling Perspective

The world of information technology is more than ever being flooded with huge amounts of data, nearly 2.5 quintillion bytes every day. This large stream of data is called big data, and the amount is increasing each day. This research uses a technique called sampling, which selects a representative subset of the data points, manipulates and analyzes this subset to identify patterns and trends in the larger dataset being examined, and finally, creates models. Sampling uses a small proportion of the original data for analysis and model training, so that it is relatively faster while maintaining data integrity and achieving accurate results. Two deep neural networks, AlexNet and DenseNet, were used in this research to test two sampling techniques, namely sampling with replacement and reservoir sampling. The dataset used for this research was divided into three classes: acceptable, flagged as easy, and flagged as hard. The base models were trained with the whole dataset, whereas the other models were trained on 50% of the original dataset. There were four combinations of model and sampling technique. The F-measure for the AlexNet model was 0.807 while that for the DenseNet model was 0.808. Combination 1 was the AlexNet model and sampling with replacement, achieving an average F-measure of 0.8852. Combination 3 was the AlexNet model and reservoir sampling. It had an average F-measure of 0.8545. Combination 2 was the DenseNet model and sampling with replacement, achieving an average F-measure of 0.8017. Finally, combination 4 was the DenseNet model and reservoir sampling. It had an average F-measure of 0.8111. Overall, we conclude that both models trained on a sampled dataset gave equal or better results compared to the base models, which used the whole dataset.     

Localized Generalization Error of Gaussian-based Classifiers and Visualization of Decision Boundaries

In pattern classification problem, one trains a classifier to recognize future unseen samples using a training dataset. Practically, one should not expect the trained classifier could correctly recognize samples dissimilar to the training dataset. Therefore, finding the generalization capability of a classifier for those unseen samples may not help in improving the classifiers accuracy. The localized generalization error model was proposed to bound above the generalization mean square error for those unseen samples similar to the training dataset only. This error model is derived based on the stochastic sensitivity measure(ST-SM)of the classifiers. We present the ST-SMS for various Gaussian based classifiers: radial basis function neural networks and support vector machine in this paper. At the end of this work, we compare the decision boundaries visualization using the training samples yielding the largest sensitivity measures and the one using support vectors in the input space.     

基于条件生成对抗网络的乳腺上皮和间质区域自动分割

乳腺病理组织图像中上皮和间质区域的自动分割对乳腺癌的诊断和治疗具有非常重要的临床意义。但是由于乳腺组织病理图像中上皮和间质区域具有高度复杂性，因此一般的分割模型很难只根据提供的分割标记来有效地训练，并对两种区域进行快速、准确的分割。为此，提出一种基于条件对抗网络（cGAN）的上皮和间质分割条件对抗网络（EPScGAN）模型。在EPScGAN中，判别器的判别机制为生成器的训练提供了一个可训练的损失函数，来更加准确地衡量出生成器网络的分割结果输出和真实标记之间的误差，从而更好地指导生成器的训练。从荷兰癌症研究所（NKI）和温哥华综合医院（VGH）两个机构提供的专家标记的乳腺病理图像数据集中随机裁剪出1 286张尺寸为512×512的图像作为实验数据集，然后将该数据集按照7：3的比例划分为训练集和测试集对EPScGAN模型进行训练和测试。结果表明，EPScGAN模型在测试集的平均交并比（mIoU）为78.12%，和其他6种流行的深度学习分割模型相比较，提出的EPScGAN具有更好的分割性能。     

一种数据增强和多模型集成的细粒度分类算法

针对解决数据缺少和单个卷积网络模型性能的限制造成细粒度分类准确率不高的问 题，提出了一种数据增强和多模型集成融合的分类算法。首先通过镜像、旋转、多尺度缩放、高 斯噪声、随机剪切和色彩增强6 种变换对CompCars 数据集进行增强处理，然后采用差异化采样 数据集的方法训练CaffeNet、VGG16 和GoogleNet 3 种差异化的网络。然后采用多重集成的方法 集成多种模型的输出结果。实验中测试网络结构在不同数据增强算法和不同模型集成下的分类结 果。模型集成的分类准确率达到94.9%，比最好的单GoogleNet 模型的分类精确率提高了9.2 个 百分点。实验结果表明该算法可以有效地提高分类的准确率。