Exploration of acquisition functions for sensor placement in DeepSensor

[DaniJonesOcean]

Issue Description: This task is focused on exploring the acquisition functions available in the DeepSensor package as part of our ongoing efforts to strategically place temperature sensors in the Great Lakes. You will go through the Acquisition Functions documentation to understand the different criteria that can inform the selection of new sensor locations.

Objective:

The aim is to better understand how these acquisition functions can guide the active learning process to optimize sensor placement. By doing so, we aim to efficiently reduce uncertainties in our representation of Great Lakes surface temperature variability.

Tasks:

• Review the section on Setting Up the Active Learning environment and ensure your development environment is consistent with the provided example.
• Experiment with various acquisition functions provided by DeepSensor, including MeanStddev, pNormStddev, OracleMAE, OracleRMSE, OracleMarginalNLL, and more.
• Analyze the behavior of each acquisition function, especially with regards to how they propose new sensor locations.
• Reflect on how each acquisition function might be particularly suited to our project based on the characteristics of the Great Lakes and our existing data.

Implementation:

• Execute the example code provided in the documentation to run the GreedyAlgorithm with different acquisition functions on the pre-trained ERA5 spatial interpolation ConvNP model.
• Generate visual representations for proposed sensor placements based on each acquisition function.
• Utilize the Sequential Greedy Algorithm and the Parallel Greedy Algorithm to see if there are variations in the recommended placements.

Deliverable:

• A short comment summarizing your findings from experimenting with each acquisition function.
• Discussion about which acquisition function(s) might be most promising for the placement of temperature sensors in the Great Lakes and the reasoning behind your choice.
• Visual outputs showcasing sensor placement proposals and acquisition function evaluations.

Resources Provided:

• DeepSensor - Acquisition Functions Documentation
• DeepSensor User Guide - Active Learning

Please feel free to ask questions or share preliminary results along the way!

[DaniJonesOcean]

Roadblock could be solved by this:

https://github.com/CIGLR-ai-lab/GreatLakes-TempSensors/issues/29

[DaniJonesOcean]

Closing to refocus - can reopen if needed