What's happening?LifeNet has learned from people who have played simple online games with openminded computers. We've implemented methods for LifeNet to learn commonsense from sensors, so that we can use modified household electronics in order to perform commonsense activity recognition. LifeNet is designed to use any human language equally well by importing the Multi-Lingual ConceptNet semantic network relationship format. LifeNet's expertise is in predicting what will happen in space and time in a typical person's life and being able to show this in human language. LifeNet currently reasons over 300,000 ConceptNet English phrase nodes, using ConceptNet links and HondaOpenMind Commonsense English stories. LifeNet will soon be able to reason over sensor data of temporally fine granularity as well. We are interested in developing further the Japanese, Korean, Portuguese, and English LifeNet's from the MultiLingual The LifeNet temporal and spatial agent model LifeNet is a tool that performs inference as a model of the common person's knowledge of the commonsense states of the world in space and time. The model is currently implemented as a combination of simple probabilistic techniques that reason over any number of time slices (Pearl's "loopy" algorithm) or any probabilistic distributions in time or 3-dimensional space (non-parametric mixtures of gaussians algorithm). We are currently working on LifeNet's ability to process sensor data and infer conceptual context from these sensors. Download LifeNetGet the most recent versions of software packages and dependencies on the downloads page. | LifeNet v0.28 | | Power spectrum calculation for audio sensor streams added. | | Equiprobable clusters of powerspecta provided by Kohonen neural network. | | Automatic "concept" identification in text streams added. | | Automatic "phenomena" identification in symbolic sensor streams added. |
| 2006 | September | 29 | | LifeNet v0.20 | N-dimensional mixtures of Gaussians inference added for reasoning over specific temporal and spatial scales. | 2005 | November | 25 | | LifeNet v0.10 | Binary Markov Random Fields implemented to reason about logical relationships. | 2004 | September | 10 |
LifeNet DevelopersJose Espinosa, Bo Morgan, Push Singh, and William Williams 
LifeNet Layout LifeNet functions as the first-person human spatial and temporal reasoning engine that incorporates data from many different sources in order to add conceptual semantic data to ConceptNet and LifeNet as well as incorporate common sense physical conceptual arrangements in space and time into LifeNet. Areas still to be designed include a social reasoning processes that can be represented as super-LifeNet engines that operate 1 layer above the basic inference and knowledge base regions shown here. This design schematic shows the minimal connectivity between the Internet human interaction data collection areas (upper-right), the sensor data aquisition and learning areas (bottom), and alsothe incorporation of existing common sense knowledge bases to the LifeNet core first-person physical relationship knowledge base. |
LifeNet Publications
|
