Researchers with the University of California Berkeley and Google Research have published a new paper detailing an AI that can remove unwanted shadows from images. The algorithm focuses on two different types of shadows -- ones from external objects and ones naturally resulting from facial features -- and works to either remove or soften them in order to maintain a natural appearance. Whereas professional images are often taken in a studio with proper lighting, the average snapshot of a person is taken'in the wild' where lighting conditions may be harsh, causing dark shadows that obscure parts of the subject's face while other parts are covered with excessive highlights. The newly developed AI is designed to address this problem by targeting those unwanted shadows and highlights, removing and softening them until a clearer subject remains. The researchers say their tool works in a'realistic and controllable way,' and it could prove useful for more than just images captured in casual settings.

Our editors have compiled this directory of the best Python books based on Amazon user reviews, rating, and ability to add business value. There are loads of free resources available online (such as Solutions Review's Data Analytics Software Buyer's Guide, visual comparison matrix, and best practices section) and those are great, but sometimes it's best to do things the old fashioned way. There are few resources that can match the in-depth, comprehensive detail of one of the best Power BI books. The editors at Solutions Review have done much of the work for you, curating this comprehensive directory of the best Python books on Amazon. Titles have been selected based on the total number and quality of reader user reviews and ability to add business value. Each of the books listed in the first section of this compilation have met a minimum criteria of 15 reviews and a 4-star-or-better ranking. Below you will find a library of titles from recognized industry analysts, experienced practitioners, and subject matter experts spanning the depths of Python coding for beginners all the way to advanced data science best practices for Python users. This compilation includes publications for practitioners of all skill levels. "Python Crash Course is the world's best-selling guide to the Python programming language. In the first half of the book, you'll learn basic programming concepts, such as variables, lists, classes, and loops, and practice writing clean code with exercises for each topic. You'll also learn how to make your programs interactive and test your code safely before adding it to a project. In the second half, you'll put your new knowledge into practice with three substantial projects: a Space Invaders-inspired arcade game, a set of data visualizations with Python's handy libraries, and a simple web app you can deploy online."

The Defense Innovation Unit is bringing together the best of commercially available artificial intelligence technology and the Defense Department's vast cache of archived medical data to teach computers how to identify cancers and other medical irregularities. The result will be new tools medical professionals can use to more accurately and more quickly identify medical issues in patients. The new DIU project, called "Predictive Health," also involves the Defense Health Agency, three private-sector businesses and the Joint Artificial Intelligence Center. The new capability directly supports the development of the JAIC's warfighter health initiative, which is working with the Defense Health Agency and the military services to field AI solutions that are aimed at transforming military health care. The JAIC is also providing the funding and adding technical expertise for the broader initiative.

Summarization of speech is a difficult problem due to the spontaneity of the flow, disfluencies, and other issues that are not usually encountered in written texts. Our work presents the first application of the BERTSum model to conversational language. We generate abstractive summaries of narrated instructional videos across a wide variety of topics, from gardening and cooking to software configuration and sports. In order to enrich the vocabulary, we use transfer learning and pretrain the model on a few large cross-domain datasets in both written and spoken English. We also do preprocessing of transcripts to restore sentence segmentation and punctuation in the output of an ASR system. The results are evaluated with ROUGE and Content-F1 scoring for the How2 and WikiHow datasets. We engage human judges to score a set of summaries randomly selected from a dataset curated from HowTo100M and YouTube. Based on blind evaluation, we achieve a level of textual fluency and utility close to that of summaries written by human content creators. The model beats current SOTA when applied to WikiHow articles that vary widely in style and topic, while showing no performance regression on the canonical CNN/DailyMail dataset. Due to the high generalizability of the model across different styles and domains, it has great potential to improve accessibility and discoverability of internet content. We envision this integrated as a feature in intelligent virtual assistants, enabling them to summarize both written and spoken instructional content upon request.

Streaming classification methods assume the number of input features is fixed and always received. But in many real-world scenarios demand is some input features are reliable while others are unreliable or inconsistent. In this paper, we propose a novel deep learning-based model called Auxiliary Network (Aux-Net), which is scalable and agile. It employs a weighted ensemble of classifiers to give a final outcome. The Aux-Net model is based on the hedging algorithm and online gradient descent. It employs a model of varying depth in an online setting using single pass learning. Aux-Net is a foundational work towards scalable neural network model for a dynamic complex environment requiring ad hoc or inconsistent input data. The efficacy of Aux-Net is shown on public dataset.

Tasks in multi-task learning often correlate, conflict, or even compete with each other. As a result, a single solution that is optimal for all tasks rarely exists. Recent papers introduced the concept of Pareto optimality to this field and directly cast multi-task learning as multi-objective optimization problems, but solutions returned by existing methods are typically finite, sparse, and discrete. We present a novel, efficient method that generates locally continuous Pareto sets and Pareto fronts, which opens up the possibility of continuous analysis of Pareto optimal solutions in machine learning problems. We scale up theoretical results in multi-objective optimization to modern machine learning problems by proposing a sample-based sparse linear system, for which standard Hessian-free solvers in machine learning can be applied. We compare our method to the state-of-the-art algorithms and demonstrate its usage of analyzing local Pareto sets on various multi-task classification and regression problems. The experimental results confirm that our algorithm reveals the primary directions in local Pareto sets for trade-off balancing, finds more solutions with different trade-offs efficiently, and scales well to tasks with millions of parameters.

Generalized linear models (GLMs) extend linear regression by generating the dependent variables through a nonlinear function of a predictor in a Reproducing Kernel Hilbert Space. Despite nonconvexity of the underlying optimization problem, the GLM-tron algorithm of Kakade et al. (2011) provably learns GLMs with guarantees of computational and statistical efficiency. We present an extension of the GLM-tron to a mirror descent or natural gradient-like setting, which we call the Reflectron. The Reflectron enjoys the same statistical guarantees as the GLM-tron for any choice of potential function $\psi$. We show that $\psi$ can be used to exploit the underlying optimization geometry and improve statistical guarantees, or to define an optimization geometry and thereby implicitly regularize the model. The implicit bias of the algorithm can be used to impose advantageous -- such as sparsity-promoting -- priors on the learned weights. Our results extend to the case of multiple outputs with or without weight sharing, and we further show that the Reflectron can be used for online learning of GLMs in the realizable or bounded noise settings. We primarily perform our analysis in continuous-time, leading to simple derivations. We subsequently prove matching guarantees for a discrete implementation. We supplement our theoretical analysis with simulations on real and synthetic datasets demonstrating the validity of our theoretical results.

Machines now perform all sorts of tasks: They clean big stores, patrol borders, and help autistic children. But will they make life better for humans?

Amazon has published videos and supplementary materials from several of its internal Machine Learning University courses.

Data is the fuel for artificial intelligence. The more data we have, the better the AI will learn and find those hidden patterns, right? We have the ability to collect LOTS of data. Consider the nearly 31 billion IoT devices producing information for machine consumption. However, lots of data does not translate into good data.