Positron Emission Tomography (PET) is a functional imaging modality widely used in neuroscience studies. To obtain meaningful quantitative results from PET images, attenuation correction is necessary during image reconstruction. For PET/MR hybrid systems, PET attenuation is challenging as Magnetic Resonance (MR) images do not reflect attenuation coefficients directly. To address this issue, we present deep neural network methods to derive the continuous attenuation coefficients for brain PET imaging from MR images. With only Dixon MR images as the network input, the existing U-net structure was adopted and analysis using forty patient data sets shows it is superior than other Dixon based methods. When both Dixon and zero echo time (ZTE) images are available, apart from stacking multiple MR images along the U-net input channels, we have proposed a new network structure to extract the features from Dixon and ZTE images independently at early layers and combine them together at later layers. Quantitative analysis based on fourteen real patient data sets demonstrates that both network approaches can perform better than the standard methods, and the proposed network structure can further reduce the PET quantification error compared to the U-net structure with multiple inputs.

We propose generative neural network methods to generate DNA sequences and tune them to have desired properties. We present three approaches: creating synthetic DNA sequences using a generative adversarial network; a DNA-based variant of the activation maximization ("deep dream") design method; and a joint procedure which combines these two approaches together. We show that these tools capture important structures of the data and, when applied to designing probes for protein binding microarrays, allow us to generate new sequences whose properties are estimated to be superior to those found in the training data. We believe that these results open the door for applying deep generative models to advance genomics research.

While bigger and deeper neural network architectures continue to advance the state-of-the-art for many computer vision tasks, real-world adoption of these networks is impeded by hardware and speed constraints. Conventional model compression methods attempt to address this problem by modifying the architecture manually or using pre-defined heuristics. Since the space of all reduced architectures is very large, modifying the architecture of a deep neural network in this way is a difficult task. In this paper, we tackle this issue by introducing a principled method for learning reduced network architectures in a data-driven way using reinforcement learning. Our approach takes a larger `teacher' network as input and outputs a compressed `student' network derived from the `teacher' network. In the first stage of our method, a recurrent policy network aggressively removes layers from the large `teacher' model. In the second stage, another recurrent policy network carefully reduces the size of each remaining layer. The resulting network is then evaluated to obtain a reward -- a score based on the accuracy and compression of the network. Our approach uses this reward signal with policy gradients to train the policies to find a locally optimal student network. Our experiments show that we can achieve compression rates of more than 10x for models such as ResNet-34 while maintaining similar performance to the input `teacher' network. We also present a valuable transfer learning result which shows that policies which are pre-trained on smaller `teacher' networks can be used to rapidly speed up training on larger `teacher' networks.

Apple last week released Turi Create, an open source package that it says will make it easy for mobile app developers to infuse machine learning into their products with just a few lines of code. "You don't have to be a machine learning expert to add recommendations, object detection, image classification, image similarity, or activity classification to your app," the company says in the GitHub description for Turi Create. From a desktop computer running macOS, Linux, or Windows, Turi Create allows users to apply several machine learning algorithms, including classifiers (like nearest neighbor, SVM, random forests); regression (logistic regression, boosted decision trees); graph analytics (PageCount, K-Core decomposition, triangle count); clustering (K-Means, DBSCAN); and topic models. The software automates the application of the algorithms to a variety of input data, including text, images, audio, video, and sensor data. Users can work with large data sets with a single machine, Apple says.

The algorithms that form the heart of machine learning have been around for decades, but computers have only recently reached the level of processing power needed to use the techniques in practical scenarios. AI programs today can learn to identify objects in images and video, translate between languages, and even master arcade and board games. In some cases, like DeepMind's AlphaGo program, the AI even exceeds top humans at its task. Artificial intelligence (AI) refers to the field of emulating intelligence in general--anything from making a convincing opponent in a video game to an automated chatbot. Machine learning is one of the biggest fields within AI, and refers only to AI programs that are designed to learn and improve at their tasks with minimal outside input from the programmer.

Software developers are quickly adopting Artificial Intelligence (AI) technologies, such as natural language understanding, sentiment analysis, speech recognition, image understanding and machine learning (ML). Across a broad range of industries and sectors, AI-infused software applications and cloud services drive innovative customer experiences, augment human capabilities and transform how we live, work and play. New tools, cloud-hosted APIs and platforms make it even easier to build such applications. Modern AI applications live at the intersection of cloud computing, data platforms and AI tools. The cloud provides a powerful foundation for elastic compute and storage, while supporting special-purpose hardware such as graphics processing units (GPUs) that accelerate demanding calculations.

New KDnuggets Cartoon looks at how AI and the new technology, such as self-driving cars, can change Christmas. Reindeer: With those new self-driving sleighs we can finally relax. This cartoon was ably drawn by Jon Carter. Here are other KDnuggets Big Data, Data Mining, and Data Science Cartoons and KDnuggets posts tagged cartoon. Cartoon: What Else Can AI Guess From Your Face? Cartoon: Future Machine Learning Class Cartoon: The First Ever Self-Driving, Deep Learning Grill Cartoon: Mother Of All Data Cartoon: Machine Learning - What They Think I Do Cartoon: the distance between Espresso and Cappuccino Cartoon: Taxes, Artificial Intelligence, and Humans Cartoon: What Happens When AI Masters the March Madness Causation or Correlation: Explaining Hill Criteria using xkcd Cartoon: Perfect Valentine's Dates Found With Data Analysis Cartoon: When Self-Driving Car Machine Learning takes you too far ... A Funny Look at Big Data and Data Science Cartoon: Thanksgiving, Big Data, and Turkey Data Science.

This tutorial was originally posted here on Ben's blog, GormAnalysis. Artificial Neural Networks are all the rage. One has to wonder if the catchy name played a role in the model's own marketing and adoption. I've seen business managers giddy to mention that their products use "Artificial Neural Networks" and "Deep Learning". Would they be so giddy to say their products use "Connected Circles Models" or "Fail and Be Penalized Machines"? But make no mistake – Artificial Neural Networks are the real deal as evident by their success in a number of applications like image recognition, natural language processing, automated trading, and autonomous cars.

One of the ways AI is and will continue t be helpful in the field of healthcare is allowing medical professionals the ability to create treatment plans as well as discovering the best suited methods for helping their patients; instead of having to battle the tread-wheel of bureaucracy, nurses and physicians can focus on doing their actual jobs. Since we are in the age of big data, patient information is becoming valuable as tech giants, such as IBM and Google, are becoming more involved in acquiring this information; therefore, companies are using AI in the field known as patient data mining in a variety of ways. The AI research branch of the company recently launched a project known as Google Deepmind Health which focuses on mining medical records with the goal of providing faster and better health services; the project can go through hundreds of thousands of medical data within minutes. Also, Google's life sciences are working on a data-collecting initiative that aims to apply some of the same algorithms used to power Goggle's search button to analyze what it is that makes a person healthy. Included in this is experimenting with technologies that monitor diseases such as a digital contact lens that might detect levels of blood sugar.

Keras is winning the world of deep learning. In this tutorial, we shall learn how to use Keras and transfer learning to produce state-of-the-art results using very small datasets. We shall provide complete training and prediction code. For this comprehensive guide, we shall be using VGG network but the techniques learned here can be used to finetune Alexnet, Inception, Resnet or any other custom network architecture. In a previous tutorial, we used 2000 images of dog and cat to get a classification accuracy of 80%.