Automatic detection of colonic polyps is still an unsolved problem due to the large variation of polyps in terms of shape, texture, size, and color, and the existence of various polyp-like mimics during colonoscopy. In this study, we apply a recent region based convolutional neural network (CNN) approach for the automatic detection of polyps in images and videos obtained from colonoscopy examinations. We use a deep-CNN model (Inception Resnet) as a transfer learning scheme in the detection system. To overcome the polyp detection obstacles and the small number of polyp images, we examine image augmentation strategies for training deep networks. We further propose two efficient post-learning methods such as, automatic false positive learning and off-line learning, both of which can be incorporated with the region based detection system for reliable polyp detection. Using the large size of colonoscopy databases, experimental results demonstrate that the suggested detection systems show better performance compared to other systems in the literature. Furthermore, we show improved detection performance using the proposed post-learning schemes for colonoscopy videos.

Thanks to the fast improvement of computation, storage and distributed computing infrastructure, ML has been evolving into more complex structured models like Deep Learning (DL), Generative Adversarial Network (GAN) and Reinforcement Learning (RL) – all using neural networks. Supervised neural networks are algorithms that can differentiate and make judgements based on image or pattern recognition, after being trained with labeled data. The concept of neural networks has been around for more than forty years, however, it was near 2014 that deep learning and neural networks began to disrupt different segments and bring us closer to passing the Turing Test. Thanks to today's data gathering capabilities, and sheer volume of said data, neural networking is one of the driving trends in successful ML execution.

MIT's Computer Science and Artificial Intelligence Lab has developed a new deep learning-based AI prediction model that can anticipate the development of breast cancer up to five years in advance. Researchers working on the product also recognized that other similar projects have often had inherent bias because they were based overwhelmingly on white patient populations, and specifically designed their own model so that it is informed by "more equitable" data that ensures it's "equally accurate for white and black women." That's key, MIT notes in a blog post, because black women are more than 42 percent more likely than white women to die from breast cancer, and one contributing factor could be that they aren't as well-served by current early detection techniques. MIT says that its work in developing this technique was aimed specifically at making the assessment of health risks of this nature more accurate for minorities, who are often not well represented in development of deep learning models. The issue of algorithmic bias is a focus of a lot of industry research and even newer products forthcoming from technology companies working on deploying AI in the field.

While today's deep learning systems are able to natively analyze video, the large file sizes of high resolution movies present unique challenges in terms of storage space and computational requirements. Sampling them into sequences of still images not only allows for real-time processing of unlimited-length videos but opens the door for creative new applications like "video ngrams." The most straightforward way to sample a video into a sequence of still images is to use a fixed-rate time-based mechanism such as one frame per second. This kind of sampling is supported natively by most tools like ffmpeg and provides a simplistic and robust workflow. At the same time, it is highly inefficient, especially for videos where there is a lot of repetition. In the case of television news, a considerable portion of the airtime is devoted to motionless anchors sitting in an unchanging studio, meaning there can be quite literally thousands of nearly identical frames in a single broadcast.

A team of MIT researchers is making it easier for novices to get their feet wet with artificial intelligence, while also helping experts advance the field. In a paper presented at the Programming Language Design and Implementation conference this week, the researchers describe a novel probabilistic-programming system named "Gen." Users write models and algorithms from multiple fields where AI techniques are applied -- such as computer vision, robotics, and statistics -- without having to deal with equations or manually write high-performance code. Gen also lets expert researchers write sophisticated models and inference algorithms -- used for prediction tasks -- that were previously infeasible. In their paper, for instance, the researchers demonstrate that a short Gen program can infer 3-D body poses, a difficult computer-vision inference task that has applications in autonomous systems, human-machine interactions, and augmented reality.

In partnership with local health insurer, CDPHP, researchers from the Institute for Data Exploration and Applications (IDEA) at Rensselaer Polytechnic Institute are using artificial intelligence to improve patient health by developing a better understanding of high needs patients and identifying aspects of care that lead to better outcomes. "It's not enough to just figure out who are the highest needs patients, you really need to know why and what approaches can help them," said Kristin Bennett, a Rensselaer math professor and associate director of IDEA. "Our approach develops explainable models that help us understand who these high needs patients are, why some people in this group do well, and some do not." The project builds on the "cadre" modeling technique developed by Bennett. As opposed to deep learning, in which a computer identifies a pattern but the path to its decision is not clear, cadre models bring another level of understanding into the equation.

I AM staring into the future of security… and it thinks I'm under 15 years old. Inside a giant hall are hundreds of stands with companies showing off their technology, from artificially intelligent video analysis to fingerprint scanners that could replace credit cards. "The words'security', 'AI' and'deep learning' are everywhere, but'privacy' is nowhere in sight" "Deep learning", "security" and "AI" are emblazoned on multiple banners throughout IFSEC, a large security conference at the ExCeL exhibition centre in London. Booths are laden with high-end security cameras, but, tellingly, the word "privacy" is nowhere in sight.

Characterization of a non-Gaussian channelized conductivity field in subsurface flow and transport modeling through inverse modeling usually leads to a high-dimensional inverse problem and requires repeated evaluations of the forward model. In this study, we develop a convolutional adversarial autoencoder (CAAE) network to parameterize the high-dimensional non-Gaussian conductivity fields using a low-dimensional latent representation and a deep residual dense convolutional network (DRDCN) to efficiently construct a surrogate model for the forward model. The two networks are both based on a multilevel residual learning architecture called residual-in-residual dense block. The multilevel residual learning strategy and the dense connection structure in the dense block ease the training of deep networks, enabling us to efficiently build deeper networks that have an essentially increased capacity for approximating mappings of very high-complexity. The CCAE and DRDCN networks are incorporated into an iterative local updating ensemble smoother to formulate an inversion framework. The integrated method is demonstrated using a synthetic solute transport model. Results indicate that CAAE is a robust parameterization method for the channelized conductivity fields with Gaussian conductivities within each facies. The DRDCN network is able to obtain an accurate surrogate model of the forward model with high-dimensional and highly-complex concentration fields using relatively limited training data. The CAAE paramterization approach and the DRDCN surrogate method together significantly reduce the number of forward model runs required to achieve accurate inversion results.

The rise in popularity of major social media platforms have enabled people to share photos and textual information about their daily life. One of the popular topics about which information is shared is food. Since a lot of media about food are attributed to particular locations and restaurants, information like popularity of spatio-temporal popularity of various cuisines can be analysed. Tracking the popularity of food types and retail locations across space and time can also be useful for business owners and restaurant investors. In this work, we present an approach using off-the shelf machine learning techniques to identify trends and popularity of cuisine types in an area using geo-tagged data from social media, Google images and Yelp. After adjusting for time, we use the Kernel Density Estimation to get hot spots across the location and model the dependencies among food cuisines popularity using Bayesian Networks. We consider the Manhattan borough of New York City as the location for our analyses but the approach can be used for any area with social media data and information about retail businesses.

We present RUSLAN -- a new open Russian spoken language corpus for the text-to-speech task. RUSLAN contains 22200 audio samples with text annotations -- more than 31 hours of high-quality speech of one person -- being the largest annotated Russian corpus in terms of speech duration for a single speaker. We trained an end-to-end neural network for the text-to-speech task on our corpus and evaluated the quality of the synthesized speech using Mean Opinion Score test. Synthesized speech achieves 4.05 score for naturalness and 3.78 score for intelligibility on a 5-point MOS scale.