For the detail of the project please see FYP_report.pdf
To run these tools, the depend packages need to be installed.
$ pip install -r requirements.txt
Notice that, it is recommend to installed in virtual environment.
These tools is developed on Python 2.7
.
For quick test, please run the command directly without any parameters.
$ python ml_algorithm/ml_algorithm.py
$ python standard_bisim/test_cases_generator.py
To run the wide and deep experiments:
$ python experiments.py
Follows are specific direction:
$ python ml_algorithm/ml_algorithm.py -h
usage: ml_algorithm.py [-h] [-e EPOCH] [-l LEARNING_RATE] [-b BATCH_SIZE]
[-p DATA_PATH] [-r TEST_TRAIN_RATE]
[-c CONTINUE_TRAIN] [-n MODEL_NAME]
optional arguments:
-h, --help show this help message and exit
-e EPOCH, --epoch EPOCH
Number of training epochs
-l LEARNING_RATE, --learning_rate LEARNING_RATE
Initial learning rate
-b BATCH_SIZE, --batch_size BATCH_SIZE
Number of data for one batch
-p DATA_PATH, --data_path DATA_PATH
Path to input data
-r TEST_TRAIN_RATE, --test_train_rate TEST_TRAIN_RATE
The rate of test cases and train cases
-c CONTINUE_TRAIN, --continue CONTINUE_TRAIN
Continue last training
-n MODEL_NAME, --model_name MODEL_NAME
The name of the model
$ python standard_bisim/test_cases_generator.py -h
usage: test_cases_generator.py [-h] [-t {random,all}] [-n NUMBER]
[-f FILE_NAME] [-v NODE_NUMBER]
[-e EDGE_TYPE_NUMBER] [-r P_RATE]
[-p PROBABILITY]
optional arguments:
-h, --help show this help message and exit
-t {random,all}, --type {random,all}
Type of data set
-n NUMBER, --number NUMBER
The length of data set
-f FILE_NAME, --file_name FILE_NAME
Name of the output file
-v NODE_NUMBER, --node_number NODE_NUMBER
Number of the nodes of the graph in the data set
-e EDGE_TYPE_NUMBER, --edge_type-number EDGE_TYPE_NUMBER
The total types of the edge in the graphs
-r P_RATE, --p_rate P_RATE
Rate of the positive cases over all cases
-p PROBABILITY, --probability PROBABILITY
The density of the random generate graphs
To visulise the performance of machine learning please use TensorBoard
$ tensorboard --logdir <path-to-summary-folder> --host localhost
Note: the experiments data in the report can be downloaded HERE