Releases: ml4ai/delphi
ASKE M5 Release
Delphi v4.0.0
This release corresponds to the ASKE month 5 milestone deliverable. A couple of bug fixes have been implemented since the alpha version.
ASKE M5 Milestone deliverable
This release corresponds to the month 5 deliverable for the DARPA Automated Scientific Knowledge Extraction program.
This release includes many enhancements and new features, most notably in the program analysis-related modules.
The API now includes - in addition to the AnalysisGraph class - the GroundedFunctionNetwork class that represents causal models assembled using information from source code.
Delphi v3.0.0
The program analysis module now correctly accounts for the path distance between a for-loop index any variables within the loop.
Delphi API documentation now has a new home: https://delphi.readthedocs.io/en/latest/
Delphi v3.0.0-alpha
Delphi now has preliminary implementations of program analysis tools to convert FORTRAN source code to Dynamic Bayes Nets.
The internal structure of the executable CAG has been changed so that arbitrary update functions can be associated with nodes rather than just updating the entire CAG by matrix multiplication.
Other features include API methods for CAG
- creation
- manipulation
- visualization
- execution
We also have updated documentation at https://delphi-framework.readthedocs.io/en/master/
Delphi v2.1.0
This release
- Adds functionality to export an 'executable' dynamic Bayes net (DBN) ensemble model in the JSON format requested by Charles River Analytics. The JSON file contains information about the link structure of the corresponding causal analysis graph, as well as discrete probability tables for the distribution of the elements of the transition matrix of the DBN.
- Reduces the size of the executable model generated for ISI by creating KDEs from a downsampled version of the gradable adjective survey data, as opposed to the original.
- Fixes a bug where the emission function was influencing the transition model when it is not supposed to.
- Gets rid of some dependencies.
- Adds improved documentation.
Delphi 2.0.0
This is a major release that shifts the focus of Delphi from CAG visualization to assembling and executing Dynamic Bayes Network Models.
Much of the visualization capabilities from the previous version of Delphi have been integrated into INDRA - you can now use INDRA and Jupyter notebooks to visualize CAGs.