What is the characteristic of Shapley Values in relation to AutoML?

Shapley Values provide a principled way of attributing the contribution of each feature in a model towards the overall prediction. They take into account all possible combinations of features, making it a thorough but computationally intensive process. This characteristic is particularly relevant in the context of AutoML, where one of the challenges is managing the computational resources effectively.

In many cases, the calculation of Shapley Values can be quite resource-intensive, leading to longer running times and higher memory usage, especially with large datasets and complex models. As a result, they are not typically computed by default in various AutoML frameworks since their cost may not align with the expectations of quick and efficient model training and evaluation.

This distinguishes Shapley Values from other techniques that might be faster to compute and more memory-efficient, as the use of Shapley Values is often reserved for scenarios where a detailed understanding of feature impact is necessary, rather than being a common automatic process.