Large Scale Model-Based Machine Learning

Model-Based Machine Learning using deterministic approximations is well suited to large scale environments, since it allows practitioners to develop sophisticated models that can still be solved efficiently. Infer.NET is a framework for performing inference on Bayesian Factor Graphs, and has been designed from the ground up to be computationally efficient. The compiler architecture means that the generated inference code often approaches the efficiency of hand-written code. Infer.NET also supports batch-processing of large datasets by sharing variables between models and the use of sparse messages when using Expectation Propagation. Customised message operators can also be implemented to overcome particular performance bottlenecks.