abstract: Generative Adversarial Networks are designed, in theory, to replicate the distribution of the data they are trained on. With real-world limitations, such as finite network capacity and training set size, they inevitably suffer a yet unavoidable technical failure: mode collapse. GAN-generated data is not nearly as diverse as the real-world data the network is trained on; this work shows that this effect is especially drastic when the training data is highly non-uniform. Specifically, GANs learn to exacerbate the social biases which exist in the training set along sensitive axes such as gender and race. In an age where many datasets are curated from web and social media data (which are almost never balanced), this has dangerous implications for downstream tasks using GAN-generated synthetic data, such as data augmentation for classification. This thesis presents an empirical demonstration of this phenomenon and illustrates its real-world ramifications. It starts by showing that when asked to sample images from an illustrative dataset of engineering faculty headshots from 47 U.S. universities, unfortunately skewed toward white males, a DCGAN’s generator “imagines” faces with light skin colors and masculine features. In addition, this work verifies that the generated distribution diverges more from the real-world distribution when the training data is non-uniform than when it is uniform. This work also shows that a conditional variant of GAN is not immune to exacerbating sensitive social biases. Finally, this work contributes a preliminary case study on Snapchat’s explosively popular GAN-enabled “My Twin” selfie lens, which consistently lightens the skin tone for women of color in an attempt to make faces more feminine. The results and discussion of the study are meant to caution machine learning practitioners who may unsuspectingly increase the biases in their applications. / Dissertation/Thesis / Masters Thesis Computer Science 2020

Deep nueral networks (DNNs) have become popular due to their predictive power and flexibility in model fitting. In unsupervised learning, variational autoencoders (VAEs) and generative adverarial networks (GANs) are two most popular and successful generative models. How to provide a unifying framework combining the best of VAEs and GANs in a principled way is a challenging task. In supervised learning, the demand for high-dimensional data analysis has grown significantly, especially in the applications of social networking, bioinformatics, and neuroscience. How to simultaneously approximate the true underlying nonlinear system and identify relevant features based on high-dimensional data (typically with the sample size smaller than the dimension, a.k.a. small-n-large-p) is another challenging task.<div><br></div><div>In this dissertation, we have provided satisfactory answers for these two challenges. In addition, we have illustrated some promising applications using modern machine learning methods.<br></div><div><br></div><div>In the first chapter, we introduce a novel inferential Wasserstein GAN (iWGAN) model, which is a principled framework to fuse auto-encoders and WGANs. GANs have been impactful on many problems and applications but suffer from unstable training. The Wasserstein GAN (WGAN) leverages the Wasserstein distance to avoid the caveats in the minmax two-player training of GANs but has other defects such as mode collapse and lack of metric to detect the convergence. The iWGAN model jointly learns an encoder network and a generator network motivated by the iterative primal dual optimization process. The encoder network maps the observed samples to the latent space and the generator network maps the samples from the latent space to the data space. We establish the generalization error bound of iWGANs to theoretically justify the performance of iWGANs. We further provide a rigorous probabilistic interpretation of our model under the framework of maximum likelihood estimation. The iWGAN, with a clear stopping criteria, has many advantages over other autoencoder GANs. The empirical experiments show that the iWGAN greatly mitigates the symptom of mode collapse, speeds up the convergence, and is able to provide a measurement of quality check for each individual sample. We illustrate the ability of iWGANs by obtaining a competitive and stable performance with state-of-the-art for benchmark datasets. <br></div><div><br></div><div>In the second chapter, we present a general framework for high-dimensional nonlinear variable selection using deep neural networks under the framework of supervised learning. The network architecture includes both a selection layer and approximation layers. The problem can be cast as a sparsity-constrained optimization with a sparse parameter in the selection layer and other parameters in the approximation layers. This problem is challenging due to the sparse constraint and the nonconvex optimization. We propose a novel algorithm, called Deep Feature Selection, to estimate both the sparse parameter and the other parameters. Theoretically, we establish the algorithm convergence and the selection consistency when the objective function has a Generalized Stable Restricted Hessian. This result provides theoretical justifications of our method and generalizes known results for high-dimensional linear variable selection. Simulations and real data analysis are conducted to demonstrate the superior performance of our method.<br></div><div><br></div><div><div>In the third chapter, we develop a novel methodology to classify the electrocardiograms (ECGs) to normal, atrial fibrillation and other cardiac dysrhythmias as defined by the Physionet Challenge 2017. More specifically, we use piecewise linear splines for the feature selection and a gradient boosting algorithm for the classifier. In the algorithm, the ECG waveform is fitted by a piecewise linear spline, and morphological features related to the piecewise linear spline coefficients are extracted. XGBoost is used to classify the morphological coefficients and heart rate variability features. The performance of the algorithm was evaluated by the PhysioNet Challenge database (3658 ECGs classified by experts). Our algorithm achieves an average F1 score of 81% for a 10-fold cross validation and also achieved 81% for F1 score on the independent testing set. This score is similar to the top 9th score (81%) in the official phase of the Physionet Challenge 2017.</div></div><div><br></div><div>In the fourth chapter, we introduce a novel region-selection penalty in the framework of image-on-scalar regression to impose sparsity of pixel values and extract active regions simultaneously. This method helps identify regions of interest (ROI) associated with certain disease, which has a great impact on public health. Our penalty combines the Smoothly Clipped Absolute Deviation (SCAD) regularization, enforcing sparsity, and the SCAD of total variation (TV) regularization, enforcing spatial contiguity, into one group, which segments contiguous spatial regions against zero-valued background. Efficient algorithm is based on the alternative direction method of multipliers (ADMM) which decomposes the non-convex problem into two iterative optimization problems with explicit solutions. Another virtue of the proposed method is that a divide and conquer learning algorithm is developed, thereby allowing scaling to large images. Several examples are presented and the experimental results are compared with other state-of-the-art approaches. <br></div>

Generative Adversarial Networks (GANs) have attracted much attention because of their ability to learn high-dimensional, realistic data distributions. In the field of medical imaging, they can be used to augment the often small image sets available. In this way, for example, the training of image classification or segmentation models can be improved to support clinical decision making. GANs can be distinguished according to their input. While Noise-to-Image GANs synthesize new images from a random noise vector, Image-To-Image GANs translate a given image into another domain. In this study, it is investigated if the performance of a Noise-To-Image GAN, defined by its generated output quality and diversity, can be improved by using elements of a previously trained Image-To-Image GAN within its training. The data used consists of paired T1- and T2-weighted MR brain images. With the objective of generating additional T1-weighted images, a hybrid model (Hybrid GAN) is implemented that combines elements of a Deep Convolutional GAN (DCGAN) as a Noise-To-Image GAN and a Pix2Pix as an Image-To-Image GAN. Thereby, starting from the dependency of an input image, the model is gradually converted into a Noise-to-Image GAN. Performance is evaluated by the use of an independent classifier that estimates the divergence between the generative output distribution and the real data distribution. When comparing the Hybrid GAN performance with the DCGAN baseline, no improvement, neither in the quality nor in the diversity of the generated images, could be observed. Consequently, it could not be shown that the performance of a Noise-To-Image GAN is improved by using elements of a previously trained Image-To-Image GAN within its training.

Machine Learning is a fast growing area that revolutionizes computer programs by providing systems with the ability to automatically learn and improve from experience. In most cases, the training process begins with extracting patterns from data. The data is a key factor for machine learning algorithms, without data the algorithms will not work. Thus, having sufficient and relevant data is crucial for the performance. In this thesis, the researcher tackles the problem of not having a sufficient dataset, in terms of the number of training examples, for an image classification task. The idea is to use Generative Adversarial Networks to generate synthetic images similar to the ground truth, and in this way expand a dataset. Two types of experiments were conducted: the first was used to fine-tune a Deep Convolutional Generative Adversarial Network for a specific dataset, while the second experiment was used to analyze how synthetic data examples affect the accuracy of a Convolutional Neural Network in a classification task. Three well known datasets were used in the first experiment, namely MNIST, Fashion-MNIST and Flower photos, while two datasets were used in the second experiment: MNIST and Fashion-MNIST. The results of the generated images of MNIST and Fashion-MNIST had good overall quality. Some classes had clear visual errors while others were indistinguishable from ground truth examples. When it comes to the Flower photos, the generated images suffered from poor visual quality. One can easily tell the synthetic images from the real ones. One reason for the bad performance is due to the large quantity of noise in the Flower photos dataset. This made it difficult for the model to spot the important features of the flowers. The results from the second experiment show that the accuracy does not increase when the two datasets, MNIST and Fashion-MNIST, are expanded with synthetic images. This is not because the generated images had bad visual quality, but because the accuracy turned out to not be highly dependent on the number of training examples. It can be concluded that Deep Convolutional Generative Adversarial Networks are capable of generating synthetic images similar to the ground truth and thus can be used to expand a dataset. However, this approach does not completely solve the initial problem of not having adequate datasets because Deep Convolutional Generative Adversarial Networks may themselves require, depending on the dataset, a large quantity of training examples.

The old methods used for images inpainting of the Depth Image Based Rendering (DIBR) process are inefficient in producing high-quality virtual views from captured data. From the viewpoint of the original image, the generated data’s structure seems less distorted in the virtual view obtained by translation but when then the virtual view involves rotation, gaps and missing spaces become visible in the DIBR generated data. The typical approaches for filling the disocclusion tend to be slow, inefficient, and inaccurate. In this project, a modern technique Generative Adversarial Network (GAN) is used to fill the disocclusion. GAN consists of two or more neural networks that compete against each other and get trained. This study result shows that GAN can inpaint the disocclusion with a consistency of the structure. Additionally, another method (Filling) is used to enhance the quality of GAN and DIBR images. The statistical evaluation of results shows that GAN and filling method enhance the quality of DIBR images.

abstract: Time series forecasting is the prediction of future data after analyzing the past data for temporal trends. This work investigates two fields of time series forecasting in the form of Stock Data Prediction and the Opioid Incident Prediction. In this thesis, the Stock Data Prediction Problem investigates methods which could predict the trends in the NYSE and NASDAQ stock markets for ten different companies, nine of which are part of the Dow Jones Industrial Average (DJIA). A novel deep learning model which uses a Generative Adversarial Network (GAN) is used to predict future data and the results are compared with the existing regression techniques like Linear, Huber, and Ridge regression and neural network models such as Long-Short Term Memory (LSTMs) models. In this thesis, the Opioid Incident Prediction Problem investigates methods which could predict the location of future opioid overdose incidences using the past opioid overdose incidences data. A similar deep learning model is used to predict the location of the future overdose incidences given the two datasets of the past incidences (Connecticut and Cincinnati Opioid incidence datasets) and compared with the existing neural network models such as Convolution LSTMs, Attention-based Convolution LSTMs, and Encoder-Decoder frameworks. Experimental results on the above-mentioned datasets for both the problems show the superiority of the proposed architectures over the standard statistical models. / Dissertation/Thesis / Masters Thesis Computer Science 2019

No description available.

No description available.

No description available.

Most semantic segmentation methods based on Convolutional Neural Networks (CNNs) rely on supervised pixel-level labelling, but because pixel-level labelling is time-consuming and laborious, synthetic images are generated by software, and their label information is already embedded inside the data; therefore, labelling can be done automatically. This advantage makes synthetic datasets widely used in training deep learning models for real-world cases. Still, compared to supervised learning with real-world labelled images, the accuracy of the models trained using synthetic datasets is not high when applied to real-world data. So, researchers have turned their interest to Unsupervised Domain Adaptation (UDA), which is mainly used to transfer knowledge learned from one domain to another. That is why we can use synthetic data to train the model. Then, the model can use what it learned to deal with real-world problems. UDA is an essential part of transfer learning. It aims to make two domain feature distributions as close as possible. In other words, UDA is mainly used to migrate the learned knowledge from one domain to another, so the knowledge and distribution learned from the source domain feature space can be migrated to the target space to improve the prediction accuracy of the target domain. However, compared with the traditional supervised learning model, the accuracy of UDA is not high when the trained UDA is used for scene segmentation of real images. The reason for the low accuracy of UDA is that the domain gap between the source and target domains is too large. The image distribution information learned by the model from the source domain cannot be applied to the target domain, which limits the development of UDA. Therefore we propose a new UDA model called TwinLossGAN, which will reduce the domain gap in two steps. The first step is to mix images from the source and target domains. The purpose is to allow the model to learn the features of images from both domains well. Mixing is performed by selecting a synthetic image on the source domain and then selecting a real-world image on the target domain. The two selected images are input to the segmenter to obtain semantic segmentation results separately. Then, the segmentation results are fed into the mixing module. The mixing model uses the ClassMix method to copy and paste some segmented objects from one image into another using segmented masks. Additionally, it generates inter-domain composite images and the corresponding pseudo-label. Then, in the second step, we modify a Generative Adversarial Network (GAN) to reduce the gap between domains further. The original GAN network has two main parts: generator and discriminator. In our proposed TwinLossGAN, the generator performs semantic segmentation on the source domain images and the target domain images separately. Segmentations are trained in parallel. The source domain synthetic images are segmented, and the loss is computed using synthetic labels. At the same time, the generated inter-domain composite images are fed to the segmentation module. The module compares its semantic segmentation results with the pseudo-label and calculates the loss. These calculated twin losses are used as generator loss for the GAN cycle for iterations. The GAN discriminator examines whether the semantic segmentation results originate from the source or target domain. The premise was that we retrieved data from GTA5 and SYNTHIA as the source domain data and images from CityScapes as the target domain data. The result was that the accuracy indicated by the TwinLossGAN that we proposed was much higher than the base UDA models.