In July, China unveiled a plan to become the world leader in artificial intelligence and create an industry worth $150 billion to its economy by 2030.

A machine learning technology developed to spot dangerous medical conditions before they worsen has shown its value in a trial. Based in Kfar Malal, Israel, Medial EarlySign is one of dozens of companies now developing Artificial Intelligence algorithms which can scan large volumes of medical records to pick out warning signs buried in patient data. Medial's latest algorithm looks to identify diabetes patients who are at highest risk of having renal dysfunction within the next 12 months. Medial EarlySign's machine learning-based model analysed dozens of factors contained in Electronic Health Records (EHRs), including laboratory test results, demographics, medication, diagnostic codes and others, to predict who might be at high risk for having renal dysfunction within the period. The company chose to isolate less than 5% of the 400,000 diabetic population selected from its database of 15 million patients, and the algorithm was able to identify 45% of patients who would progress to significant kidney damage within a year, prior to becoming symptomatic.

Deep convolutional neural networks have achieved great success in various applications. However, training an effective DNN model for a specific task is rather challenging because it requires a prior knowledge or experience to design the network architecture, repeated trial-and-error process to tune the parameters, and a large set of labeled data to train the model. In this paper, we propose to overcome these challenges by actively adapting a pre-trained model to a new task with less labeled examples. Specifically, the pre-trained model is iteratively fine tuned based on the most useful examples. The examples are actively selected based on a novel criterion, which jointly estimates the potential contribution of an instance on optimizing the feature representation as well as improving the classification model for the target task. On one hand, the pre-trained model brings plentiful information from its original task, avoiding redesign of the network architecture or training from scratch; and on the other hand, the labeling cost can be significantly reduced by active label querying. Experiments on multiple datasets and different pre-trained models demonstrate that the proposed approach can achieve cost-effective training of DNNs.

Feature missing is a serious problem in many applications, which may lead to low quality of training data and further significantly degrade the learning performance. While feature acquisition usually involves special devices or complex process, it is expensive to acquire all feature values for the whole dataset. On the other hand, features may be correlated with each other, and some values may be recovered from the others. It is thus important to decide which features are most informative for recovering the other features as well as improving the learning performance. In this paper, we try to train an effective classification model with least acquisition cost by jointly performing active feature querying and supervised matrix completion. When completing the feature matrix, a novel target function is proposed to simultaneously minimize the reconstruction error on observed entries and the supervised loss on training data. When querying the feature value, the most uncertain entry is actively selected based on the variance of previous iterations. In addition, a bi-objective optimization method is presented for cost-aware active selection when features bear different acquisition costs. The effectiveness of the proposed approach is well validated by both theoretical analysis and experimental study.

Deep convolutional network has been the state-of-the-art approach for a wide variety of tasks over the last few years. Its successes have, in many cases, turned it into the default model in quite a few domains. In this work we will demonstrate that convolutional networks have limitations that may, in some cases, hinder it from learning properties of the data, which are easily recognizable by traditional, less demanding, models. To this end, we present a series of competitive analysis studies on image recognition and text analysis tasks, for which convolutional networks are known to provide state-of-the-art results. In our studies, we inject a truth-reveling signal, indiscernible for the network, thus hitting time and again the network's blind spots. The signal does not impair the network's existing performances, but it does provide an opportunity for a significant performance boost by models that can capture it. The various forms of the carefully designed signals shed a light on the strengths and weaknesses of convolutional network, which may provide insights for both theoreticians that study the power of deep architectures, and for practitioners that consider to apply convolutional networks to the task at hand.

Scaling multinomial logistic regression to datasets with very large number of data points and classes has not been trivial. This is primarily because one needs to compute the log-partition function on every data point. This makes distributing the computation hard. In this paper, we present a distributed stochastic gradient descent based optimization method (DS-MLR) for scaling up multinomial logistic regression problems to massive scale datasets without hitting any storage constraints on the data and model parameters. Our algorithm exploits double-separability, an attractive property we observe in the objective functions of several models in machine learning, that allows us to achieve both data as well as model parallelism simultaneously. In addition to being parallelizable, our algorithm can also easily be made non-blocking and asynchronous. We demonstrate the effectiveness of DS-MLR empirically on several real-world datasets, the largest being a reddit dataset created out of 1.7 billion user comments, where the data and parameter sizes are 228 GB and 358 GB respectively.

This paper investigates estimating the variance of a temporal-difference learning agent's update target. Most reinforcement learning methods use an estimate of the value function, which captures how good it is for the agent to be in a particular state and is mathematically expressed as the expected sum of discounted future rewards (called the return). These values can be straightforwardly estimated by averaging batches of returns using Monte Carlo methods. However, if we wish to update the agent's value estimates during learning--before terminal outcomes are observed--we must use a different estimation target called the {\lambda}-return, which truncates the return with the agent's own estimate of the value function. Temporal difference learning methods estimate the expected {\lambda}-return for each state, allowing these methods to update online and incrementally, and in most cases achieve better generalization error and faster learning than Monte Carlo methods. Naturally one could attempt to estimate higher-order moments of the {\lambda}-return. This paper is about estimating the variance of the {\lambda}-return. Prior work has shown that given estimates of the variance of the {\lambda}-return, learning systems can be constructed to (1) mitigate risk in action selection, and (2) automatically adapt the parameters of the learning process itself to improve performance. Unfortunately, existing methods for estimating the variance of the {\lambda}-return are complex and not well understood empirically. We contribute a method for estimating the variance of the {\lambda}-return directly using policy evaluation methods from reinforcement learning. Our approach is significantly simpler than prior methods that independently estimate the second moment of the {\lambda}-return. Empirically our new approach behaves at least as well as existing approaches, but is generally more robust.

Britain's Home Office unveiled a tool Tuesday that algorithmically detects ISIS propaganda videos on small video hosting sites. The technology, developed by ASI data science, is designed to detect and remove videos that have been created by ISIS. Thousands of hours of ISIS video content was analyzed in order to "teach" the software's artificial intelligence what to look out for. The program looks for certain cues, but much of the proprietary information wasn't released for security reasons. The British government said that the technology was developed in order to prevent smaller video content publishers without large budgets from inadvertently spreading ISIS content.

Drone delivery is expected to take off big time in the next few years. Chinese online retailer JD.com has already launched drone delivery in four provinces in China, while DHL and Zipline are delivering medicines with drones in rural and hard-to-reach areas. Amazon, Google, and UPS are all working on getting drone delivery service off the ground. There are a lot of issues to think about when it comes to package delivery using drones--safety, privacy, and logistics being some of the main concerns. In a new study, researchers tackle two other important aspects: energy use and greenhouse gas emissions.

The rise of driverless cars and autonomous vehicles has led many to believe that the end of public transportation is nigh. From improving driving conditions to safer roads, the innovative technology promises to revolutionize how people move in modern cities. But even though driverless cars may offer relief for many of the problems plaguing individuals in cities, they will not resolve the biggest rising issue in urban transportation: gridlock. The issue of city congestion goes well beyond who's behind the wheel, whether man or machine. Having more cars on the road, even if they are autonomous, will generally lead to higher congestion.