PAC Learning of RTAs

See PAC Learning of DOTAs for a more general and powerful learning tool.

Overview

This tool is dedicated to learning real-time automata (RTAs), which is a variant of Lesliea - RTALearning based on L* algorithm. This tool for learning RTAs under more realistic assumptions within the framework of PAC learning. Membership queries and equivalence queries are realized via testing.

Installation

The project was developed using Python3, and you only need to download the project, but there are a few prerequisites before running：

Python3.7.* (or high)
graphviz (used for drawing)

Usage

Before executing the program, we need to prepare two files, model.json (including model structure) and precondition.json (including model information known to learners).

model.json

{
	"inputs": ["a", "b"],  // input actions
	"states": ["0", "1"],  // state
	"trans": {  // transition set（source state, input, timeGuards, target state）
		"0": ["0", "b", "[0,2)U[3,6)", "0"],
		"1": ["0", "a", "[5,7)", "1"],
		"2": ["1", "b", "[0,7)", "1"]
	},
	"initState": "0",  // initial target
	"acceptStates": ["1"]  // accept state
}

precondition.json

{
  "inputs": ["a", "b"], // 输入actions
  "upperGuard": 7,  // 时间上界
  "sampleNum": 6, // 采样数（符号化学习使用）
  "epsilon": 0.01, // 精确度
  "delta": 0.01 // 置信度
}

We provide model examples in Automata/Model, and users can also customize the model structure according to file model.json. In learningrta/main.py or learningrta_ symbol/main.py, users can set the target model file path, and run it directly.

learningrta/main.py: a common method of learning RTAs combined with PAC.
learningrta_ symbol/main.py: an advanced method of learning RTAs combined with symbolization method and PAC.

Reference

Maler O , Mens I E . A Generic Algorithm for Learning Symbolic Automata from Membership Queries[M]// Models, Algorithms, Logics and Tools. 2017.
Dima C. Real-time automata.[J]. Journal of Automata Languages & Combinatorics, 2001, 6(6):3-24.
Dana, Angluin. Learning regular sets from queries and counterexamples[J]. Information & Computation, 1987.

License

See MIT LICENSE file.

Contact

Please let me know if you have any questions 👉 EnvisionShen@gmail.com