Deep learning motivated by convolutional neural networks has been highly successful in a range of medical imaging problems like image classification, image segmentation, image synthesis etc. However for validation and interpretability, not only do we need the predictions made by the model but also how confident it is while making those predictions. This is important in safety critical applications for the people to accept it. In this work, we used an encoder decoder architecture based on variational inference techniques for segmenting brain tumour images. We compare different backbones architectures like U-Net, V-Net and FCN as sampling data from the conditional distribution for the encoder. We evaluate our work on the publicly available BRATS dataset using Dice Similarity Coefficient (DSC) and Intersection Over Union (IOU) as the evaluation metrics. Our model outperforms previous state of the art results while making use of uncertainty quantification in a principled bayesian manner.

In this paper, we exploit the properties of mean absolute error (MAE) as a loss function for the deep neural network (DNN) based vector-to-vector regression. The goal of this work is two-fold: (i) presenting performance bounds of MAE, and (ii) demonstrating new properties of MAE that make it more appropriate than mean squared error (MSE) as a loss function for DNN based vector-to-vector regression. First, we show that a generalized upper-bound for DNN-based vector- to-vector regression can be ensured by leveraging the known Lipschitz continuity property of MAE. Next, we derive a new generalized upper bound in the presence of additive noise. Finally, in contrast to conventional MSE commonly adopted to approximate Gaussian errors for regression, we show that MAE can be interpreted as an error modeled by Laplacian distribution. Speech enhancement experiments are conducted to corroborate our proposed theorems and validate the performance advantages of MAE over MSE for DNN based regression.

Modeling user preferences (long-term history) and user dynamics (short-term history) is of greatest importance to build efficient sequential recommender systems. The challenge lies in the successful combination of the whole user's history and his recent actions (sequential dynamics) to provide personalized recommendations. Existing methods capture the sequential dynamics of a user using fixed-order Markov chains (usually first order chains) regardless of the user, which limits both the impact of the past of the user on the recommendation and the ability to adapt its length to the user profile. In this article, we propose to use frequent sequences to identify the most relevant part of the user history for the recommendation. The most salient items are then used in a unified metric model that embeds items based on user preferences and sequential dynamics. Extensive experiments demonstrate that our method outperforms state-of-the-art, especially on sparse datasets. We show that considering sequences of varying lengths improves the recommendations and we also emphasize that these sequences provide explanations on the recommendation.

In this paper we address the instability issue of generative adversarial network (GAN) by proposing a new similarity metric in unitary space of Schur decomposition for 2D representations of audio and speech signals. We show that encoding departure from normality computed in this vector space into the generator optimization formulation helps to craft more comprehensive spectrograms. We demonstrate the effectiveness of binding this metric for enhancing stability in training with less mode collapse compared to baseline GANs. Experimental results on subsets of UrbanSound8k and Mozilla common voice datasets have shown considerable improvements on the quality of the generated samples measured by the Fr\'echet inception distance. Moreover, reconstructed signals from these samples, have achieved higher signal to noise ratio compared to regular LS-GANs.

Sensitivity to adversarial noise hinders deployment of machine learning algorithms in security-critical applications. Although many adversarial defenses have been proposed, robustness to adversarial noise remains an open problem. The most compelling defense, adversarial training, requires a substantial increase in processing time and it has been shown to overfit on the training data. In this paper, we aim to overcome these limitations by training robust models in low data regimes and transfer adversarial knowledge between different models. We train a meta-optimizer which learns to robustly optimize a model using adversarial examples and is able to transfer the knowledge learned to new models, without the need to generate new adversarial examples. Experimental results show the meta-optimizer is consistent across different architectures and data sets, suggesting it is possible to automatically patch adversarial vulnerabilities.

Deep neural networks have made breakthroughs in a wide range of visual understanding tasks. A typical challenge that hinders their real-world applications is that unknown samples may be fed into the system during the testing phase, but traditional deep neural networks will wrongly recognize these unknown samples as one of the known classes. Open set recognition (OSR) is a potential solution to overcome this problem, where the open set classifier should have the flexibility to reject unknown samples and meanwhile maintain high classification accuracy in known classes. Probabilistic generative models, such as Variational Autoencoders (VAE) and Adversarial Autoencoders (AAE), are popular methods to detect unknowns, but they cannot provide discriminative representations for known classification. In this paper, we propose a novel framework, called Conditional Probabilistic Generative Models (CPGM), for open set recognition. The core insight of our work is to add discriminative information into the probabilistic generative models, such that the proposed models can not only detect unknown samples but also classify known classes by forcing different latent features to approximate conditional Gaussian distributions. We discuss many model variants and provide comprehensive experiments to study their characteristics. Experiment results on multiple benchmark datasets reveal that the proposed method significantly outperforms the baselines and achieves new state-of-the-art performance.

Crowd Counting is a difficult but important problem in computer vision. Convolutional Neural Networks based on estimating the density map over the image has been highly successful in this domain. However dense crowd counting remains an open problem because of severe occlusion and perspective view in which people can be present at various sizes. In this work, we propose a new network which uses a ResNet based feature extractor, downsampling block which uses dilated convolutions and upsampling block using transposed convolutions. We present a novel aggregation module which makes our network robust to the perspective view problem. We present the optimization details, loss functions and the algorithm used in our work. On evaluating on ShanghaiTech, UCF-CC-50 and UCF-QNRF datasets using MSE and MAE as evaluation metrics, our network outperforms previous state of the art approaches while giving uncertainty estimates in a principled bayesian manner.

This paper presents an empirical study regarding training probabilistic neural networks using training objectives derived from PAC-Bayes bounds. In the context of probabilistic neural networks, the output of training is a probability distribution over network weights. We present two training objectives, used here for the first time in connection with training neural networks. These two training objectives are derived from tight PAC-Bayes bounds. We also re-implement a previously used training objective based on a classical PAC-Bayes bound, to compare the properties of the predictors learned using the different training objectives. We compute risk certificates that are valid on any unseen examples for the learnt predictors. We further experiment with different types of priors on the weights (both data-free and data-dependent priors) and neural network architectures. Our experiments on MNIST and CIFAR-10 show that our training methods produce competitive test set errors and non-vacuous risk bounds with much tighter values than previous results in the literature, showing promise not only to guide the learning algorithm through bounding the risk but also for model selection. These observations suggest that the methods studied here might be good candidates for self-certified learning, in the sense of certifying the risk on any unseen data without the need for data-splitting protocols.

Trail cameras are automatically triggered by animal movements. They are used by ecologists and wildlife managers around the world to study wild animal behavior and help preserve endangered species. I want to see if MATLAB image processing and deep learning can be used to identify individual animal species that visit trail cameras. We are going to start with off-the-shelf functionality--nothing specialized for this particular task. My partners on this project are Heather Gorr and Jim Sanderson. Heather is a machine learning expert at MathWorks.

A research group, consisting of astronomers mainly from the National Astronomical Observatory of Japan (NAOJ), applied a deep-learning technique, a type of AI, to classify galaxies in a large dataset of images obtained with the Subaru Telescope. Thanks to its high sensitivity, as many as 560,000 galaxies have been detected in the images. It would be extremely difficult to visually process this large number of galaxies one by one with human eyes for morphological classification. The AI enabled the team to perform the processing without human intervention. Automated processing techniques for extraction and judgment of features with deep-learning algorithms have been rapidly developed since 2012.