There are several attacks against deep learning models in the literature, including fast-gradient sign method (FGSM), basic iterative method (BIM) or momentum iterative method (MIM) attacks. These attack are the purest form of the gradient-based evading technique that is used by attackers to evade the classification model. In this work, I will present a new approach to protect a malicious activity detection model from the several adversarial machine learning attacks. Hence, we explore the power of applying adversarial training to build a robust model against FGSM attacks. Accordingly, (1) dataset enhanced with the adversarial examples; (2) deep neural network-based detection model is trained using the KDDCUP99 dataset to learn the FGSM based attack patterns.

The content of this post is a partial reproduction of a chapter from the book: "Deep Learning with PyTorch Step-by-Step: A Beginner's Guide". What do gradient descent, the learning rate, and feature scaling have in common? Every time we train a deep learning model, or any neural network for that matter, we're using gradient descent (with backpropagation). We use it to minimize a loss by updating the parameters/weights of the model. The parameter update depends on two values: a gradient and a learning rate. The learning rate gives you control of how big (or small) the updates are going to be. A bigger learning rate means bigger updates and, hopefully, a model that learns faster.

A team of researchers from Microsoft and Zhajiang University have recently created an AI model capable of singing in numerous languages. As VentureBeat reported, the DeepSinger AI developed by the team was trained on data from various music websites, using algorithms that captured the timbre of the singer's voice. Generating the "voice" of an AI singer requires algorithms that are capable of predicting and controlling both the pitch and duration of audio. When people sing, the noises they produce have vastly more complex rhythms and patterns compared to simple speech. Another problem for the team to overcome was that while there is a fair amount of speaking/speech training data available, singing training data sets are fairly rare.

Author Summary Recently, quantitative-phase imaging (QPI) has demonstrated the ability to elucidate novel parameters of cellular physiology and metabolism without the need for fluorescent staining. Here, we apply label-free, low photo-toxicity QPI to yeast cells in order to identify lipid droplets (LDs), an important organelle with key implications in human health and biofuel development. Because QPI yields low specificity, we explore the use of modern machine learning methods to rapidly identify intracellular LDs with high discriminatory power and accuracy. In recent years, machine learning has demonstrated exceptional abilities to recognize and segment objects in biomedical imaging, remote sensing, and other areas. Trained machine learning classifiers can be combined with QPI within high-throughput analysis pipelines, allowing for efficient and accurate identification and quantification of cellular components.

In this project-based course, you will use the Multiclass Neural Network module in Azure Machine Learning Studio to train a neural network to recognize handwritten digits. Microsoft Azure Machine Learning Studio is a drag-and-drop tool you can use to rapidly build and deploy machine learning models on Azure. The data used in this course is the popular MNIST data set consisting of 70,000 grayscale images of hand-written digits. You are going to deploy the trained neural network model as an Azure Web service. Azure Web Services provide an interface between an application and a Machine Learning Studio workflow scoring model.

Fastbook Reading & Discussion Session #12: Chapter 14 About: This will be a 2-hour weekly study group dedicated to going through the book, "Deep Learning for Coders with fastai and PyTorch: AI applications Without a PhD" written by the creators of the fast.ai Our Fastbook Sessions are led by Sanyam Bhutani. The authors of the course were kind enough to make an early draft available as Jupyter Notebooks. We will be going through these in the format: https://github.com/fastai/fastbook Note: To respect the license created by the authors, these sessions will not be recorded.

Disney's massive archive spans the course of nearly a century of content, which can turn any search for specific characters, scenes or on-screen objects within it into a significant undertaking. However, a team of researchers from Disney's Direct-to-Consumer & International Organization (DTCI) have built a machine learning platform to help automate the digital archival of all that content. They call it the Content Genome. The CG platform is built to populate knowledge graphs with content metadata, akin to what you see in Google results if you search for Steve Jobs (below). From there, AI applications can then leverage that data to enhance search, discovery and personalization features or as Anthony Accardo, Director of Research and Development at DTCI, told Engadget, help animators find specific shots and sequences from within Disney's archive.

Deep neural networks have shown great potential in image reconstruction problems in Euclidean space. However, many reconstruction problems involve imaging physics that are dependent on the underlying non-Euclidean geometry. In this paper, we present a new approach to learn inverse imaging that exploit the underlying geometry and physics. We first introduce a non-Euclidean encoding-decoding network that allows us to describe the unknown and measurement variables over their respective geometrical domains. We then learn the geometry-dependent physics in between the two domains by explicitly modeling it via a bipartite graph over the graphical embedding of the two geometry. We applied the presented network to reconstructing electrical activity on the heart surface from body-surface potential. In a series of generalization tasks with increasing difficulty, we demonstrated the improved ability of the presented network to generalize across geometrical changes underlying the data in comparison to its Euclidean alternatives.

The pressure of ever-increasing patient demand and budget restrictions make hospital bed management a daily challenge for clinical staff. Most critical is the efficient allocation of resource-heavy Intensive Care Unit (ICU) beds to the patients who need life support. Central to solving this problem is knowing for how long the current set of ICU patients are likely to stay in the unit. In this work, we propose a new deep learning model based on the combination of temporal convolution and pointwise (1x1) convolution, to solve the length of stay prediction task on the eICU critical care dataset. The model - which we refer to as Temporal Pointwise Convolution (TPC) - is specifically designed to mitigate for common challenges with Electronic Health Records, such as skewness, irregular sampling and missing data. In doing so, we have achieved significant performance benefits of 18-51% (metric dependent) over the commonly used Long-Short Term Memory (LSTM) network, and the multi-head self-attention network known as the Transformer.

State-of-the-art face super-resolution methods employ deep convolutional neural networks to learn a mapping between low- and high- resolution facial patterns by exploring local appearance knowledge. However, most of these methods do not well exploit facial structures and identity information, and struggle to deal with facial images that exhibit large pose variations. In this paper, we propose a novel face super-resolution method that explicitly incorporates 3D facial priors which grasp the sharp facial structures. Our work is the first to explore 3D morphable knowledge based on the fusion of parametric descriptions of face attributes (e.g., identity, facial expression, texture, illumination, and face pose). Furthermore, the priors can easily be incorporated into any network and are extremely efficient in improving the performance and accelerating the convergence speed. Firstly, a 3D face rendering branch is set up to obtain 3D priors of salient facial structures and identity knowledge. Secondly, the Spatial Attention Module is used to better exploit this hierarchical information (i.e., intensity similarity, 3D facial structure, and identity content) for the super-resolution problem. Extensive experiments demonstrate that the proposed 3D priors achieve superior face super-resolution results over the state-of-the-arts.