Releases: ccrock4t/NARS-Python
NARS Python Executable - v0.4-alpha
Version 0.4 changelog:
+ Added external configuration file; no need to recompile after config changes anymore
+ Added automated test cases for vision (MNIST dataset, supervised learning)
+ Architecture change: added Vision Buffer (spatial buffer) in place of array terms
+ Temporal module improved (now can do anticipation)
+ System now does inference on goals
+ Simplify terms in 1.) inference results, 2.) inputs
GUI Updates:
+ Big GUI speedup / cleanup; Separate GUI to run on a separate Process
+ Checkbox to filter memory (to show only statement terms)
This version was used for the experiments in A Model of Unified Perception and Cognition by Wang et al. and can be used to replicate them by running MNISTVisionTests.py
).
NARS Python Executable - v0.3-alpha
Features:
- Implemented Arrays! (Array terms, statements, and sentences)
- Allow visual inputs (type "vision: " followed by a visual image array)
- The system can now execute Operations, if they have a high desire-value
- Implemented Time projection before two-premise inference
- Implemented some non-conjunctional Conditional Syllogism (Analogy, Deduction Abduction)
- Implemented some conjunctional Conditional Syllogism (Deduction & Abduction)
- System does a better job at getting related beliefs for inference
- Bug fixes
GUI Update
- Display concept term-links
- Display concept prediction links and explanation links
- Display visualization of Judgment arrays, and scrollwheel to scale it
NARS Python Executable - v0.2-alpha
Changelog:
- GUI updates!
- Accept multi-line inputs
- Changed the Task Bag to a Buffer, which accepts Eternal inputs and the outputs of the Sensorimotor Event Buffer
- Added the Event Buffer, which is where Events are first processed by the system
- The system now accepts Goals for initial processing (but doesn't do anything with them)
- Added Immediate Inference (Negation, Conversion, Contraposition, and Product-to-Image)
- Added Compositional Inference
- Added some Temporal Inference performed on Events from Sensorimotor EventBuffer (Temporal Induction and Comparison)
- Always bug fixes
NARS Python Executable - v0.1-alpha
Details about this release:
- Buffer is a bag, which contains tasks
- Memory is a bag, which contains concepts
- In each working cycle a belief interacts with a semantically related belief
- Supports saving and loading memory (type "save" or "load")
This release SUPPORTS some aspects of NAL 1-5:
- First-order and Higher-order syllogism
- Parsing Statements as Terms
- Parsing Compound Terms
- Judgments and Questions (forward/backward inference)
- Basic sets
This release DOES NOT SUPPORT some aspects NAL 1-5, nor NAL 6+:
- Variables
- Anything temporal
- Goals
- Images or structural transformations
- Advanced inference on compound terms