Explainable deep learning models are advantageous in many situations. Prior work mostly provide unimodal explanations through post-hoc approaches not part of the original system design. Explanation mechanisms also ignore useful textual information present in images. In this paper, we propose MTXNet, an end-to-end trainable multimodal architecture to generate multimodal explanations, which focuses on the text in the image. We curate a novel dataset TextVQA-X, containing ground truth visual and multi-reference textual explanations that can be leveraged during both training and evaluation. We then quantitatively show that training with multimodal explanations complements model performance and surpasses unimodal baselines by up to 7% in CIDEr scores and 2% in IoU. More importantly, we demonstrate that the multimodal explanations are consistent with human interpretations, help justify the models' decision, and provide useful insights to help diagnose an incorrect prediction. Finally, we describe a real-world e-commerce application for using the generated multimodal explanations.

"Deep Learning"/"Deep Neural Nets" is a technological marvel that is now increasingly deployed at the cutting-edge of artificial intelligence tasks. This dramatic success of deep learning in the last few years has been hinged on an enormous amount of heuristics and it has turned out to be a serious mathematical challenge to be able to rigorously explain them. In this thesis, submitted to the Department of Applied Mathematics and Statistics, Johns Hopkins University we take several steps towards building strong theoretical foundations for these new paradigms of deep-learning. In chapter 2 we show new circuit complexity theorems for deep neural functions and prove classification theorems about these function spaces which in turn lead to exact algorithms for empirical risk minimization for depth 2 ReLU nets. We also motivate a measure of complexity of neural functions to constructively establish the existence of high-complexity neural functions. In chapter 3 we give the first algorithm which can train a ReLU gate in the realizable setting in linear time in an almost distribution free set up. In chapter 4 we give rigorous proofs towards explaining the phenomenon of autoencoders being able to do sparse-coding. In chapter 5 we give the first-of-its-kind proofs of convergence for stochastic and deterministic versions of the widely used adaptive gradient deep-learning algorithms, RMSProp and ADAM. This chapter also includes a detailed empirical study on autoencoders of the hyper-parameter values at which modern algorithms have a significant advantage over classical acceleration based methods. In the last chapter 6 we give new and improved PAC-Bayesian bounds for the risk of stochastic neural nets. This chapter also includes an experimental investigation revealing new geometric properties of the paths in weight space that are traced out by the net during the training.

The AdaBelief algorithm demonstrates superior generalization ability to the Adam algorithm by viewing the exponential moving average of observed gradients. AdaBelief is proved to have a data-dependent $O(\sqrt{T})$ regret bound when objective functions are convex, where $T$ is a time horizon. However, it remains to be an open problem on how to exploit strong convexity to further improve the convergence rate of AdaBelief. To tackle this problem, we present a novel optimization algorithm under strong convexity, called FastAdaBelief. We prove that FastAdaBelief attains a data-dependant $O(\log T)$ regret bound, which is substantially lower than AdaBelief. In addition, the theoretical analysis is validated by extensive experiments performed on open datasets (i.e., CIFAR-10 and Penn Treebank) for image classification and language modeling.

We propose a parameter efficient Bayesian layer for hierarchical convolutional Gaussian Processes that incorporates Gaussian Processes operating in Wasserstein-2 space to reliably propagate uncertainty. This directly replaces convolving Gaussian Processes with a distance-preserving affine operator on distributions. Our experiments on brain tissue-segmentation show that the resulting architecture approaches the performance of well-established deterministic segmentation algorithms (U-Net), which has never been achieved with previous hierarchical Gaussian Processes. Moreover, by applying the same segmentation model to out-of-distribution data (i.e., images with pathology such as brain tumors), we show that our uncertainty estimates result in out-of-distribution detection that outperforms the capabilities of previous Bayesian networks and reconstruction-based approaches that learn normative distributions.

This paper presents new machine learning approaches to approximate the solution of optimal stopping problems. The key idea of these methods is to use neural networks, where the hidden layers are generated randomly and only the last layer is trained, in order to approximate the continuation value. Our approaches are applicable for high dimensional problems where the existing approaches become increasingly impractical. In addition, since our approaches can be optimized using a simple linear regression, they are very easy to implement and theoretical guarantees can be provided. In Markovian examples our randomized reinforcement learning approach and in non-Markovian examples our randomized recurrent neural network approach outperform the state-of-the-art and other relevant machine learning approaches.

In modern machine learning, complex models such as deep neural networks have received increasing popularity. These complicated models are known to be able to fit noisy training data sets, while at the same time achieving small test errors. In fact, this benign overfitting phenomenon is not a unique feature of deep learning. Even for kernel methods and linear models, Belkin et al. (2018) demonstrated that interpolators on the noisy training data can still perform near optimally on the test data. A series of recent works (Belkin et al., 2019b; Muthukumar et al., 2020b; Hastie et al., 2019; Bartlett et al., 2020) theoretically studied how over-parameterization can achieve small population risk. In particular in Bartlett et al. (2020) the authors considered the setting where the data are generated from a ground-truth linear model with noise, and established a tight population risk bound for the minimum norm linear interpolator with a matching lower bound. More recently, Tsigler and Bartlett (2020) further studied benign overfitting in ridge regression, and established non-asymptotic generalization bounds for over-parametrized ridge regression. They showed that those bounds are tight for a range of regularization parameter values. Notably, these results cover arbitrary covariance structure of the data, and give a nice characterization of how the spectrum of the data covariance matrix affects the population risk in the over-parameterized regime.

In machine learning, ensemble methods [10] aim to combine hypotheses to make predictive models more robust and accurate. A weighted majority vote learning procedure is an ensemble method for classification where each voter/hypothesis is assigned a weight (i.e., its influence in the final voting). Among the most famous majority vote methods, we can cite Boosting [13], Bagging [5], or Random Forest [6]. Interestingly, most of the kernel-based classifiers, like Support Vector Machines [3, 7], can be seen as majority vote of kernel functions. Understanding when and why weighted majority votes perform better than a single hypothesis is challenging. To study the generalization abilities of such majority votes, the PAC-Bayesian framework [34, 25] offers powerful tools to obtain Probably Approximately Correct (PAC) generalization bounds. Motivated by the fact that PAC-Bayesian analyses can lead to tight bounds (see e.g., [28]), developing algorithms to minimize such bounds is an important direction (e.g., [14, 11, 15, 24]). We focus on a class of PAC-Bayesian algorithms minimizing an upper bound on the majority vote's risk called the C-Bound

The sixth generation (6G) systems are generally recognized to be established on ubiquitous Artificial Intelligence (AI) and distributed ledger such as blockchain. However, the AI training demands tremendous computing resource, which is limited in most 6G devices. Meanwhile, miners in Proof-of-Work (PoW) based blockchains devote massive computing power to block mining, and are widely criticized for the waste of computation. To address this dilemma, we propose an Evolved-Proof-of-Work (E-PoW) consensus that can integrate the matrix computations, which are widely existed in AI training, into the process of brute-force searches in the block mining. Consequently, E-PoW can connect AI learning and block mining via the multiply used common computing resource. Experimental results show that E-PoW can salvage by up to 80 percent computing power from pure block mining for parallel AI training in 6G systems.

With populations ageing, the number of people with dementia worldwide is expected to triple to 152 million by 2050. Seventy percent of cases are due to Alzheimer's disease (AD) pathology and there is a 10-20 year 'pre-clinical' period before significant cognitive decline occurs. We urgently need, cost effective, objective methods to detect AD, and other dementias, at an early stage. Risk factor modification could prevent 40% of cases and drug trials would have greater chances of success if participants are recruited at an earlier stage. Currently, detection of dementia is largely by pen and paper cognitive tests but these are time consuming and insensitive to pre-clinical phases. Specialist brain scans and body fluid biomarkers can detect the earliest stages of dementia but are too invasive or expensive for widespread use. With the advancement of technology, Artificial Intelligence (AI) shows promising results in assisting with detection of early-stage dementia. Existing AI-aided methods and potential future research directions are reviewed and discussed.

Over the last years, machine learning techniques have been applied to more and more application domains, including software engineering and, especially, software quality assurance. Important application domains have been, e.g., software defect prediction or test case selection and prioritization. The ability to predict which components in a large software system are most likely to contain the largest numbers of faults in the next release helps to better manage projects, including early estimation of possible release delays, and affordably guide corrective actions to improve the quality of the software. However, developing robust fault prediction models is a challenging task and many techniques have been proposed in the literature. Closely related to estimating defect-prone parts of a software system is the question of how to select and prioritize test cases, and indeed test case prioritization has been extensively researched as a means for reducing the time taken to discover regressions in software. In this survey, we discuss various approaches in both fault prediction and test case prioritization, also explaining how in recent studies deep learning algorithms for fault prediction help to bridge the gap between programs' semantics and fault prediction features. We also review recently proposed machine learning methods for test case prioritization (TCP), and their ability to reduce the cost of regression testing without negatively affecting fault detection capabilities.