Skip to content
naoya-i edited this page Mar 10, 2013 · 16 revisions

What is Henry N700?

Henry N700 is an open-source abductive inference engine for large-scale reasoning. Henry employs three core techniques to achieve computationally efficient First-order logic abduction: (1) lifted inference (inference on a first-order level, no need for grounding), (2) efficient search techniques developed in the Operations Research field, and (3) Cutting Plane Inference (incremental optimization technique). For more technical details, see our published papers. Also, Henry has a capability to learn an abductive score function, a function that evaluates goodness of candidate explanations. To learn a score function, Henry uses a framework of online large margin training, which is shown to have both scalability and predictive performance. Here is an overview of our manual:

Why is it called "Henry N700"?

Henry is named after John Henry which plays a key role in the popular movie "Terminator."

"You're not human. You're made of metal. Not the same metal as I am, but metal." -- John Henry

Originally, Henry was an old-fashioned artificial intelligence that played chess in the popular movie "Terminator." One day, it became "he" because it had learned lots of things from the Web every night and day. We would like to see our abductive inference engine become "he" like John Henry. That's why our reasoner is called Henry.

N700 is taken from the name of a train, where I started to develop Henry. The train was a Japanese bullet train (called "Shinkansen" in Japan) with a maximum speed of 149–186 mph.

Clone this wiki locally