1. InferPy, Probabilistic Modeling with Tensorflow Made Easy

    InferPy is a high-level Python API for probabilistic modeling built on top of Edward and Tensorflow. InferPy, which is strongly inspired by Keras, focuses on being user-friendly by using an intuitive set of abstractions that make easy to deal with complex probabilistic models. It should be seen as an interface rather than a standalone machine-learning framework. In general, InferPy has the focus on enabling flexible data processing, easy-to-code probabilistic modeling, scalable inference and robust model validation.

  2. AMIDST, a Java toolbox for scalable probabilistic machine learning

    The AMIDST Toolbox is a software for scalable probabilistic machine learning with a special focus on (massive) streaming data. The toolbox supports a flexible modeling language based on probabilistic graphical models with latent variables and temporal dependencies. The specified models can be learnt from large data sets using parallel or distributed implementations of Bayesian learning algorithms for either streaming or batch data. These algorithms are based on a flexible variational message passing scheme, which supports discrete and continuous variables from a wide range of probability distributions. AMIDST also leverages existing functionality and algorithms by interfacing to software tools such as Flink, Spark, MOA, Weka, R and HUGIN. AMIDST is an open source toolbox written in Java and available at http://www.amidsttoolbox.com under the Apache Software License 2.0.

  3. Early recognition of traffic manoeuvre intentions

    Using the AMIDST Toolbox, we prototyped models for early recognition of traffic manoeuvre intentions using object oriented graphical models. Data was collected by car on-board sensors giving rise to a large and quickly evolving data stream. This work was performed in collaboration with DAIMLER and Hugin Expert.

  4. Risk prediction in credit operations

    Using the AMIDST Toolbox, we developed a model to do risk prediction in credit operations. Data was collected continuously and reported on a monthly basis, this gives rise to a streaming data classification problem. This work has been performed in collaboration with one of our partners, the Spanish bank BCC.

  5. Real time pattern recognition in drilling logs

    Using the AMIDST Toolbox, we worked on models to detect changes in pressure measurements at the bit, and spikes in the pressure that is measured at the rig. This work has been performed in collaboration with Verdande.

  6. Crime prediction using data mining methods.

    This technology transfer project (subject to a “non-disclousure agreement”) was also made in collaboration with the company Gobile. Within this project we developed novel data mining methods for predicting the spatio-temporal occurrence of the crimes of a city based on the historical record of crimes. We started from the fact that crimes do not distribute uniformly across the city. They tend to concentrate in some areas and at some time intervals. Police forces used this knowledge when making decisions over the assignment of the scarce resources. However, the application of data mining and machine learning techniques provides a rigorous approach to deal with this information and allows making better-informed predictions and decisions.

  7. Sales Force Designing using an Artificial Intelligence based approach.

    This technology transfer project (subject to a “non-disclousure agreement”) was made in collaboration with the company Gobile. The project mainly consisted on the design of an artifical based software to address the problems in the design of a sales forece. This problem involves the solution of several interrelated problems: sizing the sales force, the problem of finding the appropiate number of salesman; salesmen location, the problem of selecting the location of each salesman in one sales covarage unit; sales territory alignment, the problem of grouping or clustering sales coverage unit into larger geographical groups; and sales resources allocation, the problem of assigning work hours of the salesman todifferent sales territories, considering also a broad set of restrictions which are normallyassociated to it.

  8. Elvira, a Java tool to construct probabilistic models-driven based decision support systems.

    The Elvira system is a Java tool to construct probabilistic models-driven based decision support systems. Elvira works with Bayesian networks and influence diagrams and it can operate with discrete, continuous and temporal variables. It has an easy to use Graphical User Interface (GUI).