Adversarial Multiscale Feature Learning Framework for Overlapping Chromosome Segmentation

Chromosome karyotype analysis is of great clinical importance in the diagnosis and treatment of diseases. Since manual analysis is highly time and effort consuming, computer-assisted automatic chromosome karyotype analysis based on images is routinely used to improve the efficiency and accuracy of the analysis. However, the strip-shaped chromosomes easily overlap each other when imaged, significantly affecting the accuracy of the subsequent analysis and hindering the development of chromosome analysis instruments. In this paper, we present an adversarial, multiscale feature learning framework to improve the accuracy and adaptability of overlapping chromosome segmentation. We first adopt the nested U-shaped network with dense skip connections as the generator to explore the optimal representation of the chromosome images by exploiting multiscale features. Then we use the conditional generative adversarial network (cGAN) to generate images similar to the original ones; the training stability of the network is enhanced by applying the least-square GAN objective. Finally, we replace the common cross-entropy loss with the advanced Lovász-Softmax loss to improve the model’s optimization and accelerate the model’s convergence. Comparing with the established algorithms, the performance of our framework is proven superior by using public datasets in eight evaluation criteria, showing its great potential in overlapping chromosome segmentation.     

基于声发射和GAN-CNN的铝合金管道法兰 连接松动泄漏检测*

面向管道法兰连接松动引起的泄漏检测需求,为解决数据样本不足和减少特征指标手动选取的繁琐环节。本文,考虑到生成性对抗网络（GAN）作为数据扩充工具,已被证明能够生成与真实数据相似的样本数据。同时,卷积神经网络（CNN）作为一种深度学习方法,为自动提取数据的特征提供了一种有效的方法。开展了基于GAN和CNN的铝合金管道法兰连接松动泄漏检测研究。首先,搭建管道泄漏标定和数据采集实验台,利用声发射技术获取不同等级的原始泄漏信号。其次,采用GAN生成样本数据扩充原始数据。同时,为了评估生成模型的性能,引入统计特评估生成质量。最后,将生成的样本数据与原始数据设置为不同训练集,基于卷积神经网络构建智能分类检测模型,应用于管道泄漏检测。同时,分类检测结果与小样本智能分类方法SVM进行了比较,实验结果表明,基于GAN和CNN构建的智能分类模型可显著提高管道法兰连接松动泄漏检测精度。     

基于多尺度空洞卷积的对抗去雾网络

雾天拍摄的图像存在颜色失真、图像细节模糊的问题,对成像设备采集到的图像质量造成了负面印象。针对雾天搜集图像存在的降质问题,提出了一种基于多尺度空洞卷积的对抗去雾网络。去雾网络的生成器由不同空洞率的卷积模块组成,结合多尺度的策略增加感受野并增强去雾效果;判别器采用多个卷积模块构成,用于区分生成的去雾图像与真实无雾图像;通过计算去雾图像和真实无雾图像之间的感知距离,优化图像的纹理结构并减少噪声信号。实验结果显示,提出算法在公开数据集上获得的峰值信噪比值为22.410 dB,结构相似性值为0.844,色差值为10.545。定量和定性评估表明,采用空洞卷积和感知损失技术设计的去雾网络能够有效地恢复图像的颜色信息和纹理结构。     

DNN Intellectual Property Extraction Using Composite Data

As state-of-the-art deep neural networks are being deployed at the core level of increasingly large numbers of AI-based products and services, the incentive for “copying them” (i.e., their intellectual property, manifested through the knowledge that is encapsulated in them) either by adversaries or commercial competitors is expected to considerably increase over time. The most efficient way to extract or steal knowledge from such networks is by querying them using a large dataset of random samples and recording their output, which is followed by the training of a student network, aiming to eventually mimic these outputs, without making any assumption about the original networks. The most effective way to protect against such a mimicking attack is to answer queries with the classification result only, omitting confidence values associated with the softmax layer. In this paper, we present a novel method for generating composite images for attacking a mentor neural network using a student model. Our method assumes no information regarding the mentor’s training dataset, architecture, or weights. Furthermore, assuming no information regarding the mentor’s softmax output values, our method successfully mimics the given neural network and is capable of stealing large portions (and sometimes all) of its encapsulated knowledge. Our student model achieved 99% relative accuracy to the protected mentor model on the Cifar-10 test set. In addition, we demonstrate that our student network (which copies the mentor) is impervious to watermarking protection methods and thus would evade being detected as a stolen model by existing dedicated techniques. Our results imply that all current neural networks are vulnerable to mimicking attacks, even if they do not divulge anything but the most basic required output, and that the student model that mimics them cannot be easily detected using currently available techniques.     

High-Dimensional Separability for One- and Few-Shot Learning

This work is driven by a practical question: corrections of Artificial Intelligence (AI) errors. These corrections should be quick and non-iterative. To solve this problem without modification of a legacy AI system, we propose special ‘external’ devices, correctors. Elementary correctors consist of two parts, a classifier that separates the situations with high risk of error from the situations in which the legacy AI system works well and a new decision that should be recommended for situations with potential errors. Input signals for the correctors can be the inputs of the legacy AI system, its internal signals, and outputs. If the intrinsic dimensionality of data is high enough then the classifiers for correction of small number of errors can be very simple. According to the blessing of dimensionality effects, even simple and robust Fisher’s discriminants can be used for one-shot learning of AI correctors. Stochastic separation theorems provide the mathematical basis for this one-short learning. However, as the number of correctors needed grows, the cluster structure of data becomes important and a new family of stochastic separation theorems is required. We refuse the classical hypothesis of the regularity of the data distribution and assume that the data can have a rich fine-grained structure with many clusters and corresponding peaks in the probability density. New stochastic separation theorems for data with fine-grained structure are formulated and proved. On the basis of these theorems, the multi-correctors for granular data are proposed. The advantages of the multi-corrector technology were demonstrated by examples of correcting errors and learning new classes of objects by a deep convolutional neural network on the CIFAR-10 dataset. The key problems of the non-classical high-dimensional data analysis are reviewed together with the basic preprocessing steps including the correlation transformation, supervised Principal Component Analysis (PCA), semi-supervised PCA, transfer component analysis, and new domain adaptation PCA.     

基于迭代Tikhonov正规化的三刺激值重建光谱方法研究

光谱图像中的反射率光谱数据维数高,且与光源、设备均无关,能够比较全面、真实、客观地描述图像中物体的颜色信息。针对三色相机的光谱图像获取系统中三维色度数据重建多维光谱数据产生的光谱信息丢失、以及伴随而生的颜色信息丢失问题,提出了迭代Tikhonov正规化的光谱重建方法。首先依据色度学理论中色度值与反射率光谱之间的关系,构建反射率光谱重建方程建立起相机所获三维色度数据与高维反射率光谱数据的映射关系;然后,通过反射率光谱重建方程的病态分析,在Moore-Penrose伪逆矩阵求解思想的基础上构建迭代Tikhonov正规化方法求解反射率光谱,并利用训练样本数据通过L-曲线方法训练获取迭代Tikhonov正规化的最优正规化参数,以有效控制并改善反射率光谱重建方程求解的病态、减少重建光谱的光谱信息丢失。实验通过选取样本数据对光谱重建方法进行验证。验证实验的结果表明所提出的光谱重建方法改善了三色相机的光谱图像获取系统中重建光谱的光谱信息丢失程度,使得重建光谱的光谱误差和色度误差较其他光谱重建方法均有明显降低。     

一种基于GAN和自适应迁移学习的样本生成方法

研究了基于生成式对抗网络(GAN)和跨域自适应迁移学习的样本生成和自动标注方法。该方法利用自适应迁移学习网络,基于已有的少量可见光图像样本集,挖掘目标在红外和可见光图像中特征内在相关性,构建自适应的转换迁移学习网络模型,生成标注好的目标图像。提出的方法解决了红外图像样本数量少且标注费时的问题,为后续多频段协同目标检测和识别获得了足够的样本数据。实验结果表明:自动标注算法对实际采集的装甲目标图像和生成的装甲目标图像各1 000张进行自动标注测试,对实际装甲目标图像的标注准确率达到95%以上,对生成的装甲目标标注准确率达到83%以上;利用真实图像和生成图像的混合数据集训练的分类器的性能和使用纯真实图像时基本一致。     

生成式对抗网络的土壤有机质高光谱估测模型

已有的土壤有机质含量估测模型大多以光谱特征波段、线性和非线性模型为基础,较少考虑通过拓展样本数据建模集来提高模型的估测能力。为进一步提高土壤有机质高光谱反演模型估测精度,提出利用生成式对抗网络(GAN)合成伪高光谱数据和有机质含量的动态估测模型。选取湖南省长沙市及周边区域的水稻田为研究对象,采集土样和实测高光谱数据(350~2 500 nm),室内化学测定有机质含量。以高光谱数据和有机质含量为基础,利用生成式对抗网络生成等量新数据, 结合原始数据建模集组成增强建模集。在GAN正式训练中,每轮训练完成后,设置4个观测点(对应增强建模集中含50,100,150和239个生成样本),动态构建交叉验证岭回归(RCV)、偏最小二乘回归(PLSR)和BP神经网络(BPNN)土壤有机质含量估测模型(分别简称GAN-RCV,GAN-PLSR和GAN-BPNN),并在相同测试集上实施模型评估。实验结果表明：(1)原始数据建模集上拟合的估测模型中,交叉验证岭回归表现最佳,决定系数(R2)和均方根误差(RMSE)分别为0.831 1和0.189 6;(2)GAN的150轮正式训练中,增强建模集上动态构建的GAN-RCV,GAN-PLSR和GAN-BPNN模型性能显著提高,具体表现为：GAN-RCV的R2取得最大值0.890 9(RMSE 0.153 7)、最小值0.850 5 (RMSE 0.18)与平均值0.868 7(RMSE 0.168 6),最大R2比建模集上拟合的RCV提高了7.2%(RMSE降低了18.9%),GAN-PLSR获得R2最大值0.855 4(RMSE 0.176 9)、最小值0.727 0 (RMSE 0.243 2)与平均值0.780 1 (RMSE 0.217 7),最大R2比建模集上拟合的PLSR提高了20.6%(RMSE降低了29.5%),GAN-BPNN表现最佳,R2取得最大值0.905 2(RMSE 0.143 3)、最小值0.801 7(RMSE 0.207 3)与平均值0.868 1(RMSE 0.168 6),最大R2比建模集上拟合的BPNN提高了30.8%(RMSE降低了44.5%);(3)随着增强建模集中生成样本数量增加,模型精度提升效果呈先升后降趋势,4个观测点中第3个观测点的模型性能提升最显著。充分的实验表明：基于GAN动态构建的有机质含量估测模型显著改善了模型预测性能。依据测试集上的评估结果,可择优使用最佳模型进行后续土壤有机质含量估测。     

A Manifold Learning Perspective on Representation Learning: Learning Decoder and Representations without an Encoder

Autoencoders are commonly used in representation learning. They consist of an encoder and a decoder, which provide a straightforward method to map n-dimensional data in input space to a lower m-dimensional representation space and back. The decoder itself defines an m-dimensional manifold in input space. Inspired by manifold learning, we showed that the decoder can be trained on its own by learning the representations of the training samples along with the decoder weights using gradient descent. A sum-of-squares loss then corresponds to optimizing the manifold to have the smallest Euclidean distance to the training samples, and similarly for other loss functions. We derived expressions for the number of samples needed to specify the encoder and decoder and showed that the decoder generally requires much fewer training samples to be well-specified compared to the encoder. We discuss the training of autoencoders in this perspective and relate it to previous work in the field that uses noisy training examples and other types of regularization. On the natural image data sets MNIST and CIFAR10, we demonstrated that the decoder is much better suited to learn a low-dimensional representation, especially when trained on small data sets. Using simulated gene regulatory data, we further showed that the decoder alone leads to better generalization and meaningful representations. Our approach of training the decoder alone facilitates representation learning even on small data sets and can lead to improved training of autoencoders. We hope that the simple analyses presented will also contribute to an improved conceptual understanding of representation learning.     

Co-Training for Visual Object Recognition Based on Self-Supervised Models Using a Cross-Entropy Regularization

Automatic recognition of visual objects using a deep learning approach has been successfully applied to multiple areas. However, deep learning techniques require a large amount of labeled data, which is usually expensive to obtain. An alternative is to use semi-supervised models, such as co-training, where multiple complementary views are combined using a small amount of labeled data. A simple way to associate views to visual objects is through the application of a degree of rotation or a type of filter. In this work, we propose a co-training model for visual object recognition using deep neural networks by adding layers of self-supervised neural networks as intermediate inputs to the views, where the views are diversified through the cross-entropy regularization of their outputs. Since the model merges the concepts of co-training and self-supervised learning by considering the differentiation of outputs, we called it Differential Self-Supervised Co-Training (DSSCo-Training). This paper presents some experiments using the DSSCo-Training model to well-known image datasets such as MNIST, CIFAR-100, and SVHN. The results indicate that the proposed model is competitive with the state-of-art models and shows an average relative improvement of 5% in accuracy for several datasets, despite its greater simplicity with respect to more recent approaches.     

Inference for Inverse Power Lomax Distribution with Progressive First-Failure Censoring

This paper investigates the statistical inference of inverse power Lomax distribution parameters under progressive first-failure censored samples. The maximum likelihood estimates (MLEs) and the asymptotic confidence intervals are derived based on the iterative procedure and asymptotic normality theory of MLEs, respectively. Bayesian estimates of the parameters under squared error loss and generalized entropy loss function are obtained using independent gamma priors. For Bayesian computation, Tierney–Kadane’s approximation method is used. In addition, the highest posterior credible intervals of the parameters are constructed based on the importance sampling procedure. A Monte Carlo simulation study is carried out to compare the behavior of various estimates developed in this paper. Finally, a real data set is analyzed for illustration purposes.     

Personal Interest Attention Graph Neural Networks for Session-Based Recommendation

Session-based recommendations aim to predict a user’s next click based on the user’s current and historical sessions, which can be applied to shopping websites and APPs. Existing session-based recommendation methods cannot accurately capture the complex transitions between items. In addition, some approaches compress sessions into a fixed representation vector without taking into account the user’s interest preferences at the current moment, thus limiting the accuracy of recommendations. Considering the diversity of items and users’ interests, a personalized interest attention graph neural network (PIA-GNN) is proposed for session-based recommendation. This approach utilizes personalized graph convolutional networks (PGNN) to capture complex transitions between items, invoking an interest-aware mechanism to activate users’ interest in different items adaptively. In addition, a self-attention layer is used to capture long-term dependencies between items when capturing users’ long-term preferences. In this paper, the cross-entropy loss is used as the objective function to train our model. We conduct rich experiments on two real datasets, and the results show that PIA-GNN outperforms existing personalized session-aware recommendation methods.     

A Time-Symmetric Formulation of Quantum Entanglement

I numerically simulate and compare the entanglement of two quanta using the conventional formulation of quantum mechanics and a time-symmetric formulation that has no collapse postulate. The experimental predictions of the two formulations are identical, but the entanglement predictions are significantly different. The time-symmetric formulation reveals an experimentally testable discrepancy in the original quantum analysis of the Hanbury Brown–Twiss experiment, suggests solutions to some parts of the nonlocality and measurement problems, fixes known time asymmetries in the conventional formulation, and answers Bell’s question “How do you convert an ’and’ into an ’or’?”     

Normalized Augmented Inverse Probability Weighting with Neural Network Predictions

The estimation of average treatment effect (ATE) as a causal parameter is carried out in two steps, where in the first step, the treatment and outcome are modeled to incorporate the potential confounders, and in the second step, the predictions are inserted into the ATE estimators such as the augmented inverse probability weighting (AIPW) estimator. Due to the concerns regarding the non-linear or unknown relationships between confounders and the treatment and outcome, there has been interest in applying non-parametric methods such as machine learning (ML) algorithms instead. Some of the literature proposes to use two separate neural networks (NNs) where there is no regularization on the network’s parameters except the stochastic gradient descent (SGD) in the NN’s optimization. Our simulations indicate that the AIPW estimator suffers extensively if no regularization is utilized. We propose the normalization of AIPW (referred to as nAIPW) which can be helpful in some scenarios. nAIPW, provably, has the same properties as AIPW, that is, the double-robustness and orthogonality properties. Further, if the first-step algorithms converge fast enough, under regulatory conditions, nAIPW will be asymptotically normal. We also compare the performance of AIPW and nAIPW in terms of the bias and variance when small to moderate L1 regularization is imposed on the NNs.     

Learning from Scarce Information: Using Synthetic Data to Classify Roman Fine Ware Pottery

In this article, we consider a version of the challenging problem of learning from datasets whose size is too limited to allow generalisation beyond the training set. To address the challenge, we propose to use a transfer learning approach whereby the model is first trained on a synthetic dataset replicating features of the original objects. In this study, the objects were smartphone photographs of near-complete Roman terra sigillata pottery vessels from the collection of the Museum of London. Taking the replicated features from published profile drawings of pottery forms allowed the integration of expert knowledge into the process through our synthetic data generator. After this first initial training the model was fine-tuned with data from photographs of real vessels. We show, through exhaustive experiments across several popular deep learning architectures, different test priors, and considering the impact of the photograph viewpoint and excessive damage to the vessels, that the proposed hybrid approach enables the creation of classifiers with appropriate generalisation performance. This performance is significantly better than that of classifiers trained exclusively on the original data, which shows the promise of the approach to alleviate the fundamental issue of learning from small datasets.     

Understanding Dilated Mathematical Relationship between Image Features and the Convolutional Neural Network’s Learnt Parameters

Deep learning, in general, was built on input data transformation and presentation, model training with parameter tuning, and recognition of new observations using the trained model. However, this came with a high computation cost due to the extensive input database and the length of time required in training. Despite the model learning its parameters from the transformed input data, no direct research has been conducted to investigate the mathematical relationship between the transformed information (i.e., features, excitation) and the model’s learnt parameters (i.e., weights). This research aims to explore a mathematical relationship between the input excitations and the weights of a trained convolutional neural network. The objective is to investigate three aspects of this assumed feature-weight relationship: (1) the mathematical relationship between the training input images’ features and the model’s learnt parameters, (2) the mathematical relationship between the images’ features of a separate test dataset and a trained model’s learnt parameters, and (3) the mathematical relationship between the difference of training and testing images’ features and the model’s learnt parameters with a separate test dataset. The paper empirically demonstrated the existence of this mathematical relationship between the test image features and the model’s learnt weights by the ANOVA analysis.     

Stabilizing stochastically-forced oscillation generators with hard excitement: a confidence-domain control approach

In this paper, noise-induced destruction of self-sustained oscillations is studied for astochastically-forced generator with hard excitement. The problem is to design a feedbackregulator that can stabilize a limit cycle of the closed-loop system and to provide arequired dispersion of the generated oscillations. The approach is based on the stochasticsensitivity function (SSF) technique and confidence domain method. A theory about thesynthesis of assigned SSF is developed. For the case when this control problem isill-posed, a regularization method is constructed. The effectiveness of the new method ofconfidence domain is demonstrated by stabilizing auto-oscillations in a randomly-forcedgenerator with hard excitement.     

基于反贝叶斯学习的WDMS光谱自动识别研究

天体光谱是天体物理学重要的研究对象,通过光谱可以获取天体的许多物理、化学参数如有效温度、金属丰度、表面重力加速度和视向速度等。白矮主序双星是一类致密的双星系统,对研究致密双星的演化特别是公共包层的演化有着重要的意义。国内外的大型巡天望远镜如美国斯隆望远镜以及中国的郭守敬望远镜,每天都产生大量光谱数据。如此海量的光谱数据无法完全用人工进行分析。因此,使用机器学习方法从海量的天体光谱中自动搜索白矮主序双星光谱,有着非常现实的意义。目前的光谱自动识别方法主要通过对已有的标签样本进行分析,通过训练得到分类器,再对未知目标进行识别。这类方法对样本的数量有明确的要求。白矮主序双星的实测光谱数量有限。若要通过有限的样本集准确学习白矮主序双星的光谱特征,不仅需要扩大样本数量,还需要提高特征提取和分类算法的精度。在前期工作中,通过机器学习等方法在海量巡天数据中识别了一批白矮主序双星的光谱,为该实验提供了数据源。使用对抗神经网络生成新的白矮主序双星光谱,扩大训练数据量至原数据集约两倍的数量,增强了分类模型的泛化能力。通过反贝叶斯学习修正损失函数,将损失函数的大小与样本的方差相关联,抑制了异常数据对模型造成的影响,提升了模型的鲁棒性,解决了由于训练样本集偏差带来的梯度消失以及训练陷入局部最优解等问题。该实验基于Tensorflow深度学习库。使用Tensorflow搭建的生成对抗网络具有较好的鲁棒性,并且封装了内部实现细节,使得算法得以更好地实现。除此之外,由Tensorflow搭建的卷积神经网络在该实验中用于分类准确度测试。实验结果表明,二维卷积神经网络能够利用卷积核有效地提取白矮主序双星的卷积特征并进行分类。基于反贝叶斯学习策略的卷积神经网络分类器在白矮主序双星原始数据及对抗神经网络生成光谱的识别任务中达到了约98.3%的准确率。该方法也可用于在巡天望远镜的海量光谱中搜索其他特殊和稀少天体如激变变星、超新星等。     

A Probabilistic Re-Intepretation of Confidence Scores in Multi-Exit Models

In this paper, we propose a new approach to train a deep neural network with multiple intermediate auxiliary classifiers, branching from it. These ‘multi-exits’ models can be used to reduce the inference time by performing early exit on the intermediate branches, if the confidence of the prediction is higher than a threshold. They rely on the assumption that not all the samples require the same amount of processing to yield a good prediction. In this paper, we propose a way to train jointly all the branches of a multi-exit model without hyper-parameters, by weighting the predictions from each branch with a trained confidence score. Each confidence score is an approximation of the real one produced by the branch, and it is calculated and regularized while training the rest of the model. We evaluate our proposal on a set of image classification benchmarks, using different neural models and early-exit stopping criteria.     

A Theoretical Perspective of the Photochemical Potential in the Spectral Performance of Photovoltaic Cells

We present a novel theoretical approach to the problem of light energy conversion in thermostated semiconductor junctions. Using the classical model of a two-level atom, we deduced formulas for the spectral response and the quantum efficiency in terms of the input photons’ non-zero chemical potential. We also calculated the spectral entropy production and the global efficiency parameter in the thermodynamic limit. The heat transferred to the thermostat results in a dissipative loss that appreciably controls the spectral quantities’ behavior and, therefore, the cell’s performance. The application of the obtained formulas to data extracted from photovoltaic cells enabled us to accurately interpolate experimental data for the spectral response and the quantum efficiency of cells based on Si-, GaAs, and CdTe, among others.