The Poisson Gamma Belief Network

arXiv.org Machine Learning

To infer a multilayer representation of high-dimensional count vectors, we propose the Poisson gamma belief network (PGBN) that factorizes each of its layers into the product of a connection weight matrix and the nonnegative real hidden units of the next layer. The PGBN's hidden layers are jointly trained with an upward-downward Gibbs sampler, each iteration of which upward samples Dirichlet distributed connection weight vectors starting from the first layer (bottom data layer), and then downward samples gamma distributed hidden units starting from the top hidden layer. The gamma-negative binomial process combined with a layer-wise training strategy allows the PGBN to infer the width of each layer given a fixed budget on the width of the first layer. The PGBN with a single hidden layer reduces to Poisson factor analysis. Example results on text analysis illustrate interesting relationships between the width of the first layer and the inferred network structure, and demonstrate that the PGBN, whose hidden units are imposed with correlated gamma priors, can add more layers to increase its performance gains over Poisson factor analysis, given the same limit on the width of the first layer.


Hierarchical Symbolic Dynamic Filtering of Streaming Non-stationary Time Series Data

arXiv.org Machine Learning

This paper proposes a hierarchical feature extractor for non-stationary streaming time series based on the concept of switching observable Markov chain models. The slow time-scale non-stationary behaviors are considered to be a mixture of quasi-stationary fast time-scale segments that are exhibited by complex dynamical systems. The idea is to model each unique stationary characteristic without a priori knowledge (e.g., number of possible unique characteristics) at a lower logical level, and capture the transitions from one low-level model to another at a higher level. In this context, the concepts in the recently developed Symbolic Dynamic Filtering (SDF) is extended, to build an online algorithm suited for handling quasi-stationary data at a lower level and a non-stationary behavior at a higher level without a priori knowledge. A key observation made in this study is that the rate of change of data likelihood seems to be a better indicator of change in data characteristics compared to the traditional methods that mostly consider data likelihood for change detection. The algorithm minimizes model complexity and captures data likelihood. Efficacy demonstration and comparative evaluation of the proposed algorithm are performed using time series data simulated from systems that exhibit nonlinear dynamics. We discuss results that show that the proposed hierarchical SDF algorithm can identify underlying features with significantly high degree of accuracy, even under very noisy conditions. Algorithm is demonstrated to perform better than the baseline Hierarchical Dirichlet Process-Hidden Markov Models (HDP-HMM). The low computational complexity of algorithm makes it suitable for on-board, real time operations.


Increasing the Interpretability of Recurrent Neural Networks Using Hidden Markov Models

arXiv.org Machine Learning

As deep neural networks continue to revolutionize various application domains, there is increasing interest in making these powerful models more understandable and interpretable, and narrowing down the causes of good and bad predictions. We focus on recurrent neural networks (RNNs), state of the art models in speech recognition and translation. Our approach to increasing interpretability is by combining an RNN with a hidden Markov model (HMM), a simpler and more transparent model. We explore various combinations of RNNs and HMMs: an HMM trained on LSTM states; a hybrid model where an HMM is trained first, then a small LSTM is given HMM state distributions and trained to fill in gaps in the HMM's performance; and a jointly trained hybrid model. We find that the LSTM and HMM learn complementary information about the features in the text.


Google creates AI that can make its own plans and envisage consequences of its actions

The Independent

Google's artificial intelligence division is developing AI that can make its own plans. DeepMind says its "Imagination-Augmented Agents" can "imagine" the possible consequences of their actions, and interpret those simulations. They can then make the right decision for what it is they want to achieve. The I.F.O. is fuelled by eight electric engines, which is able to push the flying object to an estimated top speed of about 120mph. The giant human-like robot bears a striking resemblance to the military robots starring in the movie'Avatar' and is claimed as a world first by its creators from a South Korean robotic company Waseda University's saxophonist robot WAS-5, developed by professor Atsuo Takanishi and Kaptain Rock playing one string light saber guitar perform jam session A man looks at an exhibit entitled'Mimus' a giant industrial robot which has been reprogrammed to interact with humans during a photocall at the new Design Museum in South Kensington, London Electrification Guru Dr. Wolfgang Ziebart talks about the electric Jaguar I-PACE concept SUV before it was unveiled before the Los Angeles Auto Show in Los Angeles, California, U.S The Jaguar I-PACE Concept car is the start of a new era for Jaguar.


Microsoft Creates New AI Lab to Take on Google's DeepMind

#artificialintelligence

Microsoft Corp. is setting up a new research lab focused on artificial intelligence with the goal of creating more general-purpose learning systems. The new lab, called Microsoft Research AI, will be based at the company's headquarters in Redmond, Washington, and involve more than 100 scientists from across various sub-fields of artificial intelligence research, including perception, learning, reasoning and natural language processing. The goal, said Eric Horvitz, the director of Microsoft Research Labs, is to combine these disciplines to work toward more general artificial intelligence, meaning a single system that can tackle a wide-range of tasks and problems. Such a system, for instance, might be able to both plan the best route to drive through a city and also figure out how to minimize your income tax bill, while also understanding difficult human concepts like sarcasm or gestures. This differs from so-called narrow AIs, which are just designed to perform a single task well -- for instance, recognize faces in digital photographs.