2.5 KiB
Keyboard And Mouse Interactive Dataset
Neural Network
Requirements
The code is test in Ubuntu 20.04.
pytorch 1.11.0 matplotlib 3.3.2 pickle 4.0 pandas 1.4.3
Train
Set training parameters in train.sh
Run sh train.sh to train the model
Test
Run sh test.sh to run test on trained model
Predictions are saved under prediction/task$i$/
Bayesian Inference
Requirements
Run sh sampler_user.sh to split prediction to 10% to 90%
Run Rscript stan/strategy_inference_test.R to get results of intention prediction for all users
Run sh stan/plot_user.sh to plot the bar chart for user intention prediction results of all action sequences
Run Rscript stan/strategy_inference_test_full_length.R to get results of intention prediction (0% to 100%) for all users
Run sh stan/plot_user_length_10_steps.sh to plot the bar chart for user intention prediction results (0% to 100%) of all action sequences
Run sh sampler_single_act.sh to get the predictions for each individual action sequence.
Run Rscript stan/strategy_inference_test_all_individual_act.R to get all action sequences (0% to 100%) of all users for intention prediction
Run sh plot_user_all_individual.sh to plot the bar chart for user intention prediction results of all action sequences
Run sh plot_user_length_10_steps_all_individual.sh to plot the user intention prediction results (0% to 100%) of all action sequences
Set training and test parameters in train.sh and test.sh
Run sh train.sh to train the model.
Run sh test.sh to run test on trained model. Predictions are saved under prediction/task$i$/
Run sh sampler_user.sh to split prediction to 10% to 90%
Run stan/strategy_inference_test.R to get results of intention prediction for all users Run stan/plot_user.py to plot the bar chart for user intention prediction results of all action sequences
Run stan/strategy_inference_test_full_length.R to get results of intention prediction (0% to 100%) for all users Run stan/plot_user_length_10_users.py to plot the bar chart for user intention prediction results (0% to 100%) of all action sequences
Run stan/strategy_inference_test_all_individual_act.R to get all action sequences (0% to 100%) of all users for intention prediction Run stan/plot_user_all_individual.py to plot the bar chart for user intention prediction results of all action sequences Run stan/plot_user_length_10_steps_all_individual.py to plot the user intention prediction results (0% to 100%) of all action sequences