This demonstration, crafted with the collaboration of Riddhi Pitliya Jain, showcases an innovative approach to understanding emotional responses through active inference. It revolves around a scenario where we observe an agent's emotional states as they navigate through the challenge of locating their wallet within their home. This setup serves as a practical application for our emotional inference model, a sophisticated hierarchical active inference framework. At its core, the model integrates a dual-level analysis: the primary level focuses on the agent's task-specific actions (in this case, searching for a wallet), while the superior level reflecting its hierarchical nature, is tasked with the emotional analysis of the activities at the lower tier.
Additionally, the demonstration includes a control scenario—designated as the wallet-finding task devoid of the emotional inference architecture. This setup is intended to provide a comparative analysis against the comprehensive emotional-wallet-finding task, allowing for a clearer understanding of the emotional inference model's impact and effectiveness.
To initiate the experimental process, specific preparatory steps are required:
-
Begin by purging the "experiments" folder of any existing files and removing the "results.txt" file to ensure a clean slate for new experimental data.
-
Navigate to the "config.yaml" file and modify the parameters according to the needs of the upcoming experiments. This step is crucial for tailoring the experimental conditions.
-
With the preparations complete, proceed to the directory containing your experimental files within the code terminal. Execute the command ./sweep.sh config.yaml 30, substituting "30" with the desired number of experiments to conduct in a single batch. This command initiates the automated experimental process.
-
As experiments are conducted, individual results will be systematically stored within the "experiments" folder. Upon completion of the batch, a consolidated compilation of results will be generated and saved in the "results.txt" file, facilitating an organized analysis of outcomes.