Large datasets of sub-meter aerial imagery represented as orthophoto mosaics are widely available today, and these data sets may hold a great deal of untapped information. This imagery has a potential to locate several types of features; for example, forests, parking lots, airports, residential areas, or freeways in the imagery. However, the appearances of these things vary based on many things including the time that the image is captured, the sensor settings, processing done to rectify the image, and the geographical and cultural context of the region captured by the image. This thesis explores the use of deep convolutional neural networks to classify land use from very high spatial resolution (VHR), orthorectified, visible band multispectral imagery. Recent technological and commercial applications have driven the collection a massive amount of VHR images in the visible red, green, blue (RGB) spectral bands, this work explores the potential for deep learning algorithms to exploit this imagery for automatic land use/ land cover (LULC) classification.

In the last two decades, the landscape of text generation has undergone tremendous changes and is being reshaped by the success of deep learning. New technologies for text generation ranging from template-based methods to neural network-based methods emerged. Meanwhile, the research objectives have also changed from generating smooth and coherent sentences to infusing personalized traits to enrich the diversification of newly generated content. With the rapid development of text generation solutions, one comprehensive survey is urgent to summarize the achievements and track the state of the arts. In this survey paper, we present the general systematical framework, illustrate the widely utilized models and summarize the classic applications of text generation.

Applying neural-networks on Question Answering has gained increasing popularity in recent years. In this paper, I implemented a model with Bi-directional attention flow layer, connected with a Multi-layer LSTM encoder, connected with one start-index decoder and one conditioning end-index decoder. I introduce a new end-index decoder layer, conditioning on start-index output. The Experiment shows this has increased model performance by 15.16%. For prediction, I proposed a new smart-span equation, rewarding both short answer length and high probability in start-index and end-index, which further improved the prediction accuracy. The best single model achieves an F1 score of 73.97% and EM score of 64.95% on test set.

Convolutional Neural Networks are very efficient at processing signals defined on a discrete Euclidean space (such as images). However, as they can not be used on signals defined on an arbitrary graph, other models have emerged, aiming to extend its properties. We propose to review some of the major deep learning models designed to exploit the underlying graph structure of signals. We express them in a unified formalism, giving them a new and comparative reading.

We empirically investigate the (negative) expected accuracy as an alternative loss function to cross entropy (negative log likelihood) for classification tasks. Coupled with softmax activation, it has small derivatives over most of its domain, and is therefore hard to optimize. A modified, leaky version is evaluated on a variety of classification tasks, including digit recognition, image classification, sequence tagging and tree tagging, using a variety of neural architectures such as logistic regression, multilayer perceptron, CNN, LSTM and Tree-LSTM. We show that it yields comparable or better accuracy compared to cross entropy. Furthermore, the proposed objective is shown to be more robust to label noise.

A common problem in high energy physics is extracting a signal from a much larger background. Posed as a classification task, there is said to be an imbalance in the number of samples belonging to the signal class versus the number of samples from the background class. Techniques for learning from imbalanced data are well established in the machine learning community. In this work we provide a brief overview of class imbalance techniques in a high energy physics setting. Two case studies are presented: (1) the measurement of the longitudinal polarization fraction in same-sign $WW$ scattering, and (2) the decay of the Higgs boson to charm-quark pairs. We find a significant improvement in the performance of the machine learning models used in the longitudinal $WW$ study, while no significant improvement in performance is found in the deep learning models tested. Our charm-quark tagger gives a 14% improvement in the background rejection rate.

We introduce a novel loss function for training deep learning architectures to perform classification. It consists in minimizing the smoothness of label signals on similarity graphs built at the output of the architecture. Equivalently, it can be seen as maximizing the distances between the network function images of training inputs from distinct classes. As such, only distances between pairs of examples in distinct classes are taken into account in the process, and the training does not prevent inputs from the same class to be mapped to distant locations in the output domain. We show that this loss leads to similar performance in classification as architectures trained using the classical cross-entropy, while offering interesting degrees of freedom and properties. We also demonstrate the interest of the proposed loss to increase robustness of trained architectures to deviations of the inputs.

Abstract--In this work we describe the preparation of a time series dataset of inertial measurements for determining the surface type under a wheeled robot. The data consists of over 7600 labeled time series samples, with the corresponding surface type annotation. This data was used in two public competitions with over 1500 participant in total. Additionally, we describe the performance of state-of-art deep learning models for time series classification, as well as propose a baseline model based on an ensemble of machine learning methods. The baseline achieves an accuracy of over 68% with our nine-category dataset.

Deep neural networks are vulnerable against adversarial examples. In this paper, we propose to train and test the networks with randomly subsampled images with high drop rates. We show that this approach significantly improves robustness against adversarial examples in all cases of bounded L0, L2 and L_inf perturbations, while reducing the standard accuracy by a small value. We argue that subsampling pixels can be thought to provide a set of robust features for the input image and, thus, improves robustness without performing adversarial training.

Deep neural networks are vulnerable to adversarial examples, i.e., carefully-perturbed inputs aimed to mislead classification. This work proposes a detection method based on combining non-linear dimensionality reduction and density estimation techniques. Our empirical findings show that the proposed approach is able to effectively detect adversarial examples crafted by non-adaptive attackers, i.e., not specifically tuned to bypass the detection method. Given our promising results, we plan to extend our analysis to adaptive attackers in future work.