pomegranate | Fast and Flexible Probabilistic Modeling in Python | SciPy 2017 | Jacob Schreiber скачать в хорошем качестве

pomegranate | Fast and Flexible Probabilistic Modeling in Python | SciPy 2017 | Jacob Schreiber

pomegranate is a python package that extends the ideas behind scikit-learn to probabilistic models such as mixtures, Bayesian networks, and hidden Markov models. pomegranate is built to be modular by separating out the probability distributions from the models themselves, allowing both arbitrary distributions to be used for any model, and making the modeling of different features with different distributions a breeze. pomegranate was built with large quantities of data in mind, supporting both an out-of-core API for when your data doesn't fit in memory, built-in multi-threaded parallelism made possibly by releasing the GIL, and most recently GPU support for large Gaussian models. This talk will be an overview of the most important features behind pomegranate and motivate the concept of probabilistic modeling.