Prototype-Based Methods in Explainable AI and Emerging Opportunities in the Geosciences

TL;DR

This work discusses a series of developments in the field of prototype-based XAI that show potential for scientific learning tasks, with a focus on the geosciences.

Abstract

Prototype-based methods are intrinsically interpretable XAI methods that produce predictions and explanations by comparing input data with a set of learned prototypical examples that are representative of the training data. In this work, we discuss a series of developments in the field of prototype-based XAI that show potential for scientific learning tasks, with a focus on the geosciences. We organize the prototype-based XAI literature into three themes: the development and visualization of prototypes, types of prototypes, and the use of prototypes in various learning tasks. We discuss how the authors use prototype-based methods, their novel contributions, and any limitations or challenges that may arise when adapting these methods for geoscientific learning tasks. We highlight differences between geoscientific data sets and the standard benchmarks used to develop XAI methods, and discuss how specific geoscientific applications may benefit from using or modifying existing prototype-based XAI techniques.

Figures (4)

• Figure 1: Diagram of the prototype-based XAI model li2018deep with encoder-decoder architecture and classifier for an MNIST example.
• Figure 2: Notional diagram of deformable prototypes donnelly2022deformable where the prototypical patches (bounding boxes) chen2019looks vary in their spatial organization for one input image (see Fig. 5 in original paper donnelly2022deformable).
• Figure 3: Notional diagram of support prototypes found closer to the classification boundary and trivial prototypes found far from each other (see Fig. 1 in original paper wang2023learning).
• Figure 4: Notional representation of ProtoLNet's barnes2022looks prototype-based reasoning for an example location-relevant learning task classifying which quadrant the object is located. We find those prototypes with the highest similarity to the test image. The location scaling grid shows where each prototype is important for this classification.)