Despite their successes in the field of self-learning AI, Convolutional Neural Networks (CNNs) suffer from having too many trainable parameters, impacting computational performance. Several approaches have been proposed to reduce the number of parameters in the visual domain, the Inception architecture [Szegedy et al., 2016] being a prominent example. This raises the question whether the number of trainable parameters in CNNs can also be reduced for 1D inputs, such as time-series data, without incurring a substantial loss in classification performance. We propose and examine two methods for complexity reduction in AstroNet [Shallue & Vanderburg, 2018], a CNN for automatic classification of time-varying brightness data of stars to detect exoplanets. The first method makes only a tactical reduction of layers in AstroNet while the second method also modifies the original input data by means of a Gaussian pyramid. We conducted our experiments with various degrees of dropout regularization. Our results show only a non-substantial loss in accuracy compared to the original AstroNet, while reducing training time up to 85 percent. These results show potential for similar reductions in other CNN applications while largely retaining accuracy.

We consider a communication scenario, in which an intruder, employing a deep neural network (DNN), tries to determine the modulation scheme of the intercepted signal. Our aim is to minimize the accuracy of the intruder, while guaranteeing that the intended receiver can still recover the underlying message with the highest reliability. This is achieved by constellation perturbation at the encoder, similarly to adversarial attacks against DNN-based classifiers. In the latter perturbation is limited to be imperceptible to a human observer, while in our case perturbation is constrained so that the message can still be reliably decoded by the legitimate receiver which is oblivious to the perturbation. Simulation results demonstrate the viability of our approach to make wireless communication secure against DNN-based intruders with minimal sacrifice in the communication performance.

We propose a novel data-driven method to learn multiple kernels in kernel methods of statistical machine learning from training samples. The proposed kernel learning algorithm is based on a $U$-statistics of the empirical marginal distributions of features in the feature space given their class labels. We prove the consistency of the $U$-statistic estimate using the empirical distributions for kernel learning. In particular, we show that the empirical estimate of $U$-statistic converges to its population value with respect to all admissible distributions as the number of the training samples increase. We also prove the sample optimality of the estimate by establishing a minimax lower bound via Fano's method. In addition, we establish the generalization bounds of the proposed kernel learning approach by computing novel upper bounds on the Rademacher and Gaussian complexities using the concentration of measures for the quadratic matrix forms.We apply the proposed kernel learning approach to classification of the real-world data-sets using the kernel SVM and compare the results with $5$-fold cross-validation for the kernel model selection problem. We also apply the proposed kernel learning approach to devise novel architectures for the semantic segmentation of biomedical images. The proposed segmentation networks are suited for training on small data-sets and employ new mechanisms to generate representations from input images.

Over-parametrization of deep neural networks has recently been shown to be key to their successful training. However, it also renders them prone to overfitting and makes them expensive to store and train. Tensor regression networks significantly reduce the number of effective parameters in deep neural networks while retaining accuracy and the ease of training. They replace the flattening and fully-connected layers with a tensor regression layer, where the regression weights are expressed through the factors of a low-rank tensor decomposition. In this paper, to further improve tensor regression networks, we propose a novel stochastic rank-regularization. It consists of a novel randomized tensor sketching method to approximate the weights of tensor regression layers. We theoretically and empirically establish the link between our proposed stochastic rank-regularization and the dropout on low-rank tensor regression. Extensive experimental results with both synthetic data and real world datasets (i.e., CIFAR-100 and the UK Biobank brain MRI dataset) support that the proposed approach i) improves performance in both classification and regression tasks, ii) decreases overfitting, iii) leads to more stable training and iv) improves robustness to adversarial attacks and random noise.

We provide a summary over architectural approaches that can be used to construct dependable learning-enabled autonomous systems, with a focus on automated driving. We consider three technology pillars for architecting dependable autonomy, namely diverse redundancy, information fusion, and runtime monitoring. For learning-enabled components, we additionally summarize recent architectural approaches to increase the dependability beyond standard convolutional neural networks. We conclude the study with a list of promising research directions addressing the challenges of existing approaches.

We study the problem of meta-learning through the lens of online convex optimization, developing a meta-algorithm bridging the gap between popular gradient-based meta-learning and classical regularization-based multi-task transfer methods. Our method is the first to simultaneously satisfy good sample efficiency guarantees in the convex setting, with generalization bounds that improve with task-similarity, while also being computationally scalable to modern deep learning architectures and the many-task setting. Despite its simplicity, the algorithm matches, up to a constant factor, a lower bound on the performance of any such parameter-transfer method under natural task similarity assumptions. We use experiments in both convex and deep learning settings to verify and demonstrate the applicability of our theory.

We introduce a new method to tag Multiword Expressions (MWEs) using a linguistically interpretable language-independent deep learning architecture. We specifically target discontinuity, an under-explored aspect that poses a significant challenge to computational treatment of MWEs. Two neural architectures are explored: Graph Convolutional Network (GCN) and multi-head self-attention. GCN leverages dependency parse information, and self-attention attends to long-range relations. We finally propose a combined model that integrates complementary information from both through a gating mechanism. The experiments on a standard multilingual dataset for verbal MWEs show that our model outperforms the baselines not only in the case of discontinuous MWEs but also in overall F-score.

Datasets with missing values are very common on industry applications, and they can have a negative impact on machine learning models. Recent studies introduced solutions to the problem of imputing missing values based on deep generative models. Previous experiments with Generative Adversarial Networks and Variational Autoencoders showed interesting results in this domain, but it is not clear which method is preferable for different use cases. The goal of this work is twofold: we present a comparison between missing data imputation solutions based on deep generative models, and we propose improvements over those methodologies. We run our experiments using known real life datasets with different characteristics, removing values at random and reconstructing them with several imputation techniques. Our results show that the presence or absence of categorical variables can alter the selection of the best model, and that some models are more stable than others after similar runs with different random number generator seeds.

Financial time series prediction, especially with machine learning techniques, is an extensive field of study. In recent times, deep learning methods (especially time series analysis) have performed outstandingly for various industrial problems, with better prediction than machine learning methods. Moreover, many researchers have used deep learning methods to predict financial time series with various models in recent years. In this paper, we will compare various deep learning models, such as multilayer perceptron (MLP), one-dimensional convolutional neural networks (1D CNN), stacked long short-term memory (stacked LSTM), attention networks, and weighted attention networks for financial time series prediction. In particular, attention LSTM is not only used for prediction, but also for visualizing intermediate outputs to analyze the reason of prediction; therefore, we will show an example for understanding the model prediction intuitively with attention vectors. In addition, we focus on time and factors, which lead to an easy understanding of why certain trends are predicted when accessing a given time series table. We also modify the loss functions of the attention models with weighted categorical cross entropy; our proposed model produces a 0.76 hit ratio, which is superior to those of other methods for predicting the trends of the KOSPI 200.

Semi-supervised Generative Adversarial Networks (GANs) are developed in the context of travel mode inference with uni-dimensional smartphone trajectory data. We use data from a large-scale smartphone travel survey in Montreal, Canada. We convert GPS trajectories into fixed-sized segments with five channels (variables). We develop different GANs architectures and compare their prediction results with Convolutional Neural Networks (CNNs). The best semi-supervised GANs model led to a prediction accuracy of 83.4%, while the best CNN model was able to achieve the prediction accuracy of 81.3%. The results compare favorably with previous studies, especially when taking the large-scale real-world nature of the dataset into account.