New research suggests that computer models could help doctors achieve greater accuracy in the diagnosis of cancer and other diseases. A research team from Beth Israel Deaconess Medical Center (BIDMC) and Harvard Medical School (HMS) have developed an artificial intelligence (AI) system which is able to train computers to analyse pathologic image data [PDF]. The scientists hope that the programme could one day aid in diagnosing disease. 'Our AI method is based on deep learning, a machine-learning algorithm used for a range of applications including speech recognition and image recognition,' explained Andrew Beck, director of bioinformatics at the Cancer Research Institute at BIDMC and associate professor at HMS. He added: 'This approach teaches machines to interpret the complex patterns and structure observed in real-life data by building multi-layer artificial neural networks, in a process which is thought to show similarities with the learning process that occurs in layers of neurons in the brain's neocortex, the region where thinking occurs.'

Hi, I'm planning to make a 5/6 part series to reflect about my experience in University of Washington Machine Learning Specialization in Coursera while I take the five courses: Foundations, Regression, Classification, Clustering, Deep Learning and finish the capstone project. This is the first article in the series. I'll feature the first course Machine Learning Foundations: A Case Study Approach in this article and describe the philosophy behind the'case study approach' with a brief overview of the tools used and reflect on what I've learnt. I hope it will help people who want to use the same specialization.I'm also taking courses in Udacity, Edx and using other resources too, but experience those resources will be described in separate articles. I'm also planning to write a whole separate series on Udacity Machine Learning Nanodegree in recent future.

Beleaguered enterprises are struggling to keep pace with cyber threats, and small and medium-sized businesses are hit hardest of all due to limited resources. A recent survey by the Federation of Small Business (FSB) found 66% of those questioned had been a victim of cybercrime over the past two years, and only 4% had an incident response plan in place in anticipation of an attack. For many, cyber security takes them into unfamiliar territory and depletes the time spent on core business activities. This has seen an over-reliance upon point solutions, poor attention to patching and updates, and a failure to apply strategic business-specific security controls. To make matters worse, the potential attack surface is only set to widen as the Internet of Things sees sensors and IP-enabled tech insinuate themselves into every niche of society, even the small business.

The ability to represent complex high dimensional probability distributions in a compact form is one of the key insights in the field of graphical models. Factored representations are ubiquitous in machine learning and lead to major computational advantages. We explore a different type of compact representation based on discrete Fourier representations, complementing the classical approach based on conditional independencies. We show that a large class of probabilistic graphical models have a compact Fourier representation. This theoretical result opens up an entirely new way of approximating a probability distribution. We demonstrate the significance of this approach by applying it to the variable elimination algorithm. Compared with the traditional bucket representation and other approximate inference algorithms, we obtain significant improvements.

We investigate the use of Deep Neural Networks for the classification of image datasets where texture features are important for generating class-conditional discriminative representations. To this end, we first derive the size of the feature space for some standard textural features extracted from the input dataset and then use the theory of Vapnik-Chervonenkis dimension to show that hand-crafted feature extraction creates low-dimensional representations which help in reducing the overall excess error rate. As a corollary to this analysis, we derive for the first time upper bounds on the VC dimension of Convolutional Neural Network as well as Dropout and Dropconnect networks and the relation between excess error rate of Dropout and Dropconnect networks. The concept of intrinsic dimension is used to validate the intuition that texture-based datasets are inherently higher dimensional as compared to handwritten digits or other object recognition datasets and hence more difficult to be shattered by neural networks. We then derive the mean distance from the centroid to the nearest and farthest sampling points in an n-dimensional manifold and show that the Relative Contrast of the sample data vanishes as dimensionality of the underlying vector space tends to infinity.

The lasso and related sparsity inducing algorithms have been the target of substantial theoretical and applied research. Correspondingly, many results are known about their behavior for a fixed or optimally chosen tuning parameter specified up to unknown constants. In practice, however, this oracle tuning parameter is inaccessible so one must use the data to select one. Common statistical practice is to use a variant of cross-validation for this task. However, little is known about the theoretical properties of the resulting predictions with such data-dependent methods. We consider the high-dimensional setting with random design wherein the number of predictors $p$ grows with the number of observations $n$. Under typical assumptions on the data generating process, similar to those in the literature, we recover oracle rates up to a log factor when choosing the tuning parameter with cross-validation. Under weaker conditions, when the true model is not necessarily linear, we show that the lasso remains risk consistent relative to its linear oracle. We also generalize these results to the group lasso and square-root lasso and investigate the predictive and model selection performance of cross-validation via simulation.

Gaussian random fields are a powerful tool for modeling environmental processes. For high dimensional samples, classical approaches for estimating the covariance parameters require highly challenging and massive computations, such as the evaluation of the Cholesky factorization or solving linear systems. Recently, Anitescu, Chen and Stein \cite{M.Anitescu} proposed a fast and scalable algorithm which does not need such burdensome computations. The main focus of this article is to study the asymptotic behavior of the algorithm of Anitescu et al. (ACS) for regular and irregular grids in the increasing domain setting. Consistency, minimax optimality and asymptotic normality of this algorithm are proved under mild differentiability conditions on the covariance function. Despite the fact that ACS's method entails a non-concave maximization, our results hold for any stationary point of the objective function. A numerical study is presented to evaluate the efficiency of this algorithm for large data sets.

With laptops and hand-held devices slowly replacing pencils and paper, some educators question the importance of teaching handwriting in the classroom. Although some say it is a nonessential motor skill, researchers have found evidence that in fact it helps children pay attention to and understand the written language. Brain scans in children who did not yet know how to print revealed they are unable to distinguish letters and respond to them'the same as to a triangle'. Although some say it is a nonessential motor skill, researchers have found evidence this skill helps children pay attention to and understand the written language. Brain scans in children who did not yet know how to print yet revealed they are unable to distinguish letters and respond to them'the same as to a triangle' You have to see letters in'the mind's eye' in order to create them on a piece of paper she explained.

Great hardware needs great software. To help data scientists and developers make the most of the vast opportunities in deep learning, we're announcing today at the International Supercomputing show, ISC16, a trio of new capabilities for our deep learning software platform. The three -- NVIDIA DIGITS 4, CUDA Deep Neural Network Library (cuDNN) 5.1 and the new GPU Inference Engine (GIE) -- are powerful tools that make it even easier to create solutions on our platform. NVIDIA DIGITS 4 introduces a new object detection workflow, enabling data scientists to train deep neural networks to find faces, pedestrians, traffic signs, vehicles and other objects in a sea of images. This workflow enables advanced deep learning solutions -- such as tracking objects from satellite imagery, security and surveillance, advanced driver assistance systems and medical diagnostic screening.

Learning needs to be less like memorization, and more like…Angry Birds. Half of school dropouts name boredom as the number one reason they left. The post is about why the future of education will be about flipping our current model on its head and about how key exponential technologies like AI, VR and gamification are going to drive a revolution in education. In the traditional education system, you start at an "A," and every time you get something wrong, your score gets lower and lower. You start with zero, and every time you come up with something right, your score gets higher and higher. It completely flips the way we currently learn, and it's addictively fun.