This repository contains the code used for the experimentation shown in the paper presented at the 27th Pacific-Asia Conference on Knowledge Discovery and Data Mining.
Link to the paper: https://link.springer.com/chapter/10.1007/978-3-031-33383-5_26
execute:
conda create -n cpnn python=3.8
conda activate cpnn
pip install -r requirements.txt
The project is composed by the following directories.
It contains the generated data streams.
- sine_rw_10_13: S1+, S1-.
- sine_rw_10_31: S1-, S1+.
- sine_rw_10_24: S2+, S1-.
- sine_rw_10_42: S2-, S2+.
- sine_rw_10_1234: S1+, S2+, S1-, S2-.
- sine_rw_10_1432: S1+, S2-, S1-, S2+.
- sine_rw_10_2143: S2+, S1+, S2-, S1-.
- sine_rw_10_2341: S2+, S1-, S2-, S1+.
It contains Jupyter Notebooks.
It contains the python modules implementing cPNN and cLSTM.
It contains the script to execute to replicate results and plot metrics.
It contains the python modules implementing the data stream generator.
It runs the prequential evaluation. Run python -m run_test.test_then_train --help to see the arguments.
Change the variable dataset in the code to set the data stream.
It stores pickle files containing the results in the performance/<data_stream_name>/<model_name>_<hidden_size>hs folder.
All the following scripts/notebooks require the prequential evaluation to be executed on all the three architectures (cPNN, cLSTM, cLSTMs)
It plots the metrics trend of different concepts.
You should change the variable dataset in the code to set the data stream.
It stores the plots of accuracy, Kappa Temporal and Cohen Kappa in files: performance/<data_stream_name>/_plot_clstm_<hidden_size>hs/test_then_train___clstm_<metric>.png.
It performs the Shapiro-Wilk test to check for the normality of the samples.
It stores the results in performance/<data_stream_name>/_prob_test/normality_<data_stream_name> in .csv and .xlsx format.
It performs all the mentioned tests in Section 5.
It stores the results in performance/<data_stream_name>/_prob_test/test_complete_<data_stream_name> in .csv and .xlsx format.
It generates a new data stream containing concepts S1+, S2+, S1-, S2- and stores it in the dataset folder.
https://github.com/AndreaCossu/ContinualLearning-SequentialProcessing