It is the classic defence for many men being asked to do two things at once – they can't multi-task. But the excuse that it would be easier for a woman to do it is probably right, as it has emerged men's brains work far harder to juggle tasks. Researchers at the Higher School of Economics in Moscow believe this is because hunter-gatherer men evolved to better judge spatial awareness. Women's brains, on the other hand, had to evolve to cope with managing the family at home, making it less hard work to use the same skills now. Researchers at the Higher School of Economics in Moscow believe men can't multi-task because as hunter-gatherers they evolved to better judge spatial awareness Brain scans of 140 people aged 20 to 65 given the same task-switching challenge found the front of the brain lit up far more in men than women, while parts of the male brain not even needed by women got involved.

Across the globe, energy systems are changing and creating unprecedented challenges for the organisations tasked with ensuring the lights stay on. In the UK, National Grid is facing shrinking margins, looming capacity shortages and unpredictable peaks and troughs in energy supply caused by increasing levels of renewable penetration. At the Reinventing Energy Summit, Michael Bironneau, Head of Technology Development at Open Energi, will explore how the same machine learning techniques that have let machines defeat chess and Go masters, can also be leveraged to orchestrate massive amounts of flexible demand-side capacity – from industrial equipment, co-generation and battery storage systems – towards the one goal of creating a smarter grid; one that is cleaner, cheaper, more secure and more efficient. For World Cities Day 2016, I asked Michael a few questions to learn more about utilising data science in energy, creating a smarter grid, political challenges, and more. What are the main transformative technologies that will help create a smarter grid?

BellHawk Systems Corporation announces the availability of a new white paper "Artificial Intelligence and the Smart Industrial Warehouse." This white paper is available for download from the front page News section of www.BellHawk.com. In the popular press, there is much ado made about Artificial Intelligence (AI) being used in robots that buzz about high volume retail warehouses, automatically picking consumer products that are being shipped overnight in response to orders made over the Internet. But this misses all the ways that AI can be used to inexpensively improve the operation of industrial warehouses without a major investment in robots or other expensive materials handling equipment by providing the information and advice that managers, supervisors, and material handlers need to do their jobs efficiently. This white paper examines how AI based operations tracking and management systems, such as BellHawk, can be used to improve the efficiency of industrial warehouses, prevent mistakes, and enable customer orders to be shipped on time.

Machine learning technology in neural networks has been pushing artificial intelligence (AI) development to new heights. Most AI systems learn to do things using a set of labelled data provided by their human programmers. Parham Aarabi and Wenzhi Guo, engineers from the University of Toronto, Canada have taken machine learning to a different level, developing an algorithm that can learn things on its own, going beyond its training. In a study published in the journal IEEE Transactions on Neural Networks and Learning Systems, the researchers discuss how they created an algorithm that identifies people's hair in photographs. Instead of the usual method of training neural networks (exposure to existing sets of examples), their algorithm learned directly from human instructions, a model called "heuristic training."

In lexicon-based classification, documents are assigned labels by comparing the number of words that appear from two opposed lexicons, such as positive and negative sentiment. Creating such words lists is often easier than labeling instances, and they can be debugged by non-experts if classification performance is unsatisfactory. However, there is little analysis or justification of this classification heuristic. This paper describes a set of assumptions that can be used to derive a probabilistic justification for lexicon-based classification, as well as an analysis of its expected accuracy. One key assumption behind lexicon-based classification is that all words in each lexicon are equally predictive. This is rarely true in practice, which is why lexicon-based approaches are usually outperformed by supervised classifiers that learn distinct weights on each word from labeled instances. This paper shows that it is possible to learn such weights without labeled data, by leveraging co-occurrence statistics across the lexicons.

Survival analysis is a fundamental tool in medical research to identify predictors of adverse events and develop systems for clinical decision support. In order to leverage large amounts of patient data, efficient optimisation routines are paramount. We propose an efficient training algorithm for the kernel survival support vector machine (SSVM). We directly optimise the primal objective function and employ truncated Newton optimisation and order statistic trees to significantly lower computational costs compared to previous training algorithms, which require $O(n^4)$ space and $O(p n^6)$ time for datasets with $n$ samples and $p$ features. Our results demonstrate that our proposed optimisation scheme allows analysing data of a much larger scale with no loss in prediction performance. Experiments on synthetic and 5 real-world datasets show that our technique outperforms existing kernel SSVM formulations if the amount of right censoring is high ($\geq85\%$), and performs comparably otherwise.

We propose a new method to study the internal memory used by reinforcement learning policies. We estimate the amount of relevant past information by estimating mutual information between behavior histories and the current action of an agent. We perform this estimation in the passive setting, that is, we do not intervene but merely observe the natural behavior of the agent. Moreover, we provide a theoretical justification for our approach by showing that it yields an implementation-independent lower bound on the minimal memory capacity of any agent that implement the observed policy. We demonstrate our approach by estimating the use of memory of DQN policies on concatenated Atari frames, demonstrating sharply different use of memory across 49 games. The study of memory as information that flows from the past to the current action opens avenues to understand and improve successful reinforcement learning algorithms.

Many prediction problems can be phrased as inferences over local neighborhoods of graphs. The graph represents the interaction between entities, and the neighborhood of each entity contains information that allows the inferences or predictions. We present an approach for applying machine learning directly to such graph neighborhoods, yielding predicitons for graph nodes on the basis of the structure of their local neighborhood and the features of the nodes in it. Our approach allows predictions to be learned directly from examples, bypassing the step of creating and tuning an inference model or summarizing the neighborhoods via a fixed set of hand-crafted features. The approach is based on a multi-level architecture built from Long Short-Term Memory neural nets (LSTMs); the LSTMs learn how to summarize the neighborhood from data. We demonstrate the effectiveness of the proposed technique on a synthetic example and on real-world data related to crowdsourced grading, Bitcoin transactions, and Wikipedia edit reversions.

Existing methods for structure discovery in time series data construct interpretable, compositional kernels for Gaussian process regression models. While the learned Gaussian process model provides posterior mean and variance estimates, typically the structure is learned via a greedy optimization procedure. This restricts the space of possible solutions and leads to over-confident uncertainty estimates. We introduce a fully Bayesian approach, inferring a full posterior over structures, which more reliably captures the uncertainty of the model.

Over the years, ensemble methods have become a staple of machine learning. Similarly, generalized linear models (GLMs) have become very popular for a wide variety of statistical inference tasks. The former have been shown to enhance out- of-sample predictive power and the latter possess easy interpretability. Recently, ensembles of GLMs have been proposed as a possibility. On the downside, this approach loses the interpretability that GLMs possess. We show that minimum description length (MDL)-motivated compression of the inferred ensembles can be used to recover interpretability without much, if any, downside to performance and illustrate on a number of standard classification data sets.