State-space models (SSMs) are a highly expressive model class for learning patterns in time series data and for system identification. Deterministic versions of SSMs (e.g. LSTMs) proved extremely successful in modeling complex time series data. Fully probabilistic SSMs, however, are often found hard to train, even for smaller problems. To overcome this limitation, we propose a novel model formulation and a scalable training algorithm based on doubly stochastic variational inference and Gaussian processes. In contrast to existing work, the proposed variational approximation allows one to fully capture the latent state temporal correlations. These correlations are the key to robust training. The effectiveness of the proposed PR-SSM is evaluated on a set of real-world benchmark datasets in comparison to state-of-the-art probabilistic model learning methods. Scalability and robustness are demonstrated on a high dimensional problem.

Convolutional neural networks have been extremely successful in the image recognition domain because they ensure equivariance to translations. There have been many recent attempts to generalize this framework to other domains, including graphs and data lying on manifolds. In this paper we give a rigorous, theoretical treatment of convolution and equivariance in neural networks with respect to not just translations, but the action of any compact group. Our main result is to prove that (given some natural constraints) convolutional structure is not just a sufficient, but also a necessary condition for equivariance to the action of a compact group. Our exposition makes use of concepts from representation theory and noncommutative harmonic analysis and derives new generalized convolution formulae.

With the success of deep learning, recent efforts have been focused on analyzing how learned networks make their classifications. We are interested in analyzing the network output based on the network structure and information flow through the network layers. We contribute an algorithm for 1) analyzing a deep network to find neurons that are "important" in terms of the network classification outcome, and 2) automatically labeling the patches of the input image that activate these important neurons. We propose several measures of importance for neurons and demonstrate that our technique can be used to gain insight into, and explain how a network decomposes an image to make its final classification.

Bayesian learning has been recently considered as an effective means of accounting for uncertainty in trained deep network parameters. This is of crucial importance when dealing with small or sparse training datasets. On the other hand, shallow models that compute weighted sums of their inputs, after passing them through a bank of arbitrary randomized nonlinearities, have been recently shown to enjoy good test error bounds that depend on the number of nonlinearities. Inspired from these advances, in this paper we examine novel deep network architectures, where each layer comprises a bank of arbitrary nonlinearities, linearly combined using multiple alternative sets of weights. We effect model training by means of approximate inference based on a t-divergence measure; this generalizes the Kullback-Leibler divergence in the context of the t-exponential family of distributions. We adopt the t-exponential family since it can more flexibly accommodate real-world data, that entail outliers and distributions with fat tails, compared to conventional Gaussian model assumptions. We extensively evaluate our approach using several challenging benchmarks, and provide comparative results to related state-of-the-art techniques.

In this work, we propose a simple but effective method to interpret black-box machine learning models globally. That is, we use a compact binary tree, the interpretation tree, to explicitly represent the most important decision rules that are implicitly contained in the black-box machine learning models. This tree is learned from the contribution matrix which consists of the contributions of input variables to predicted scores for each single prediction. To generate the interpretation tree, a unified process recursively partitions the input variable space by maximizing the difference in the average contribution of the split variable between the divided spaces. We demonstrate the effectiveness of our method in diagnosing machine learning models on multiple tasks. Also, it is useful for new knowledge discovery as such insights are not easily identifiable when only looking at single predictions. In general, our work makes it easier and more efficient for human beings to understand machine learning models.

Montreal-based Element AI has compiled a report and analysis on the global supply of AI researchers in an effort to get a better understanding of an industry in high demand. Overall, the report found that there are 22,064 PhD-educated researchers globally that are capable of working in AI research and applications, with only 3,074 candidates currently looking for work. The US had the highest concentration of researchers with 9,010 researchers, followed by the UK with 1,861 researchers. Canada fell in third place with 1,154 researchers. To conduct the broader survey, Element AI used results from LinkedIn searches that showed the total number of profiles according to specialized parameters.

There are interesting parallels between one of this week's milestones in the history of technology and the current excitement and anxiety about artificial intelligence (AI). Bottom line: Beware of fake AI news and be less afraid. On February 10, 1996, IBM's Deep Blue became the first machine to win a chess game against a reigning world champion, Garry Kasparov. Kasparov won three and drew two of the following five games, defeating Deep Blue by a score of 4–2. In May 1997, an upgraded version of Deep Blue won the six-game rematch 3½–2½ to become the first computer to defeat a reigning world champion in a match under standard chess tournament time controls. Deep Blue was an example of so-called "artificial intelligence" achieved through "brute force," the super-human calculating speed that has been the hallmark of digital computers since they were invented in the 1940s.

Google, Microsoft, IBM and AWS are just some of the tech behemoths taking on machine learning, creating APIs and developing a number of sophisticated deep learning frameworks. As new areas of technology are exploited and pulled into the mainstream, the demand for skilled workers begins to rise. This is our guide to getting started with machine learning. First coined in 1959 by Arthur Samuel - a computer scientist at IBM at the time - "Machine Learning" essentially enables computers to learn without being directly programmed. Machine learning (ML) is fundamentally the application of AI that we recognise today, for example, machines performing'smart' tasks.

Machine learning is poised to help marketers garner phenomenal new insights and results, and to change many processes and jobs along the way. We discussed this potential in "Machine Learning is About to Turn the Marketing World Upside Down." Machine learning can't provide better results alone, of course. Marketers need to collaborate with data scientists to identify important questions to explore, accelerate tests, improve the accuracy of answers, and make better decisions. And to effectively collaborate, they need a common language.

Bidirectional LSTMs are an extension of traditional LSTMs that can improve model performance on sequence classification problems. In problems where all timesteps of the input sequence are available, Bidirectional LSTMs train two instead of one LSTMs on the input sequence. The first on the input sequence as-is and the second on a reversed copy of the input sequence. This can provide additional context to the network and result in faster and even fuller learning on the problem. In this tutorial, you will discover how to develop Bidirectional LSTMs for sequence classification in Python with the Keras deep learning library.