Open Source Linked Data, Machine Learning & Measuring the Information in Data
This was the final meetup before the main Connected Data 2017 conference and we hosted 3 great speakers!
Domenico Corapi of Grakn.AI
Khalil Ahmed of BrightstarDB
James Hoskisson of Valtech
Domenico Corapi gave us an overview of the machine learning techniques that represent training data and output as a structured knowledge base. Two experiments were shown using grakn and learning rules to train a deep neural network that complements an existing knowledge base.
Khalil Ahmed gave an overview of the opensource BrightstarDB project where he is lead developer. He provided an overview of the RDF triplestore's architecture, future developments and cutting edge ideas for the new linked data functionality.
James' talk was "How to Measure the Information in your Data”. The synopsis is:
In a world where everything is monitored and logged by many different methods, the volumes of data concerned can soon become overwhelming. Even with the dramatic increase in technologies for distributed processing and storage there is still a need to slim down the data to just the essentials. This talk introduces the concept of information entropy and how you can use information theory techniques to craft better features, reject spurious data, and reduce noise related errors in model predictions.