accumulated local effects
In plain wordsA chart of how a model's predictions change with one input, built from small local comparisons so it stays reliable when inputs are related to each other, unlike simpler average-based charts.
A method for plotting the main effect of a feature on a model's predictions that remains valid when features are correlated. The feature's range is divided into intervals, typically at quantiles; within each interval, the change in prediction as the feature moves from the interval's lower to its upper edge is computed only for instances that actually fall in that interval and averaged, and these local differences are accumulated across intervals and centered so that the effect averages to zero. Because each comparison uses realistic data points and small changes, ALE avoids the extrapolation to impossible feature combinations that biases partial dependencepartial dependence plot, n.: A chart showing how a model's average prediction changes as one input is varied while everything else stays as it is in the data — for example, predicted house price against floor area. under correlation, and it separates a feature's effect from those of correlated features. Second-order ALE plots show pure two-way interactions. Results depend on the number of intervals, and effects are conditional on the data within each one.
Unlike the partial dependence plot, the ALE plot showed no effect of bedroom count once floor area was accounted for.
Origin
Introduced by Daniel W. Apley and Jingyu Zhu in 'Visualizing the Effects of Predictor Variables in Black Box Supervised Learning Models' (arXiv, 2016; Journal of the Royal Statistical Society, Series B, 2020).
Related
- partial dependence plotpartial dependence plot, n.: A chart showing how a model's average prediction changes as one input is varied while everything else stays as it is in the data — for example, predicted house price against floor area.
- individual conditional expectationindividual conditional expectation, n.: A chart that traces, for each example separately, what a model would predict as one input is swept across its range, exposing differences between examples that an overall average would hide.
- global explanationglobal explanation, n.: An explanation of how an AI model behaves overall — which factors matter most across all its decisions, and in what way — rather than why it made one particular decision.
- feature importancefeature importance, n.: A score for how much each input — age, income, location — matters to a model's predictions. It helps people understand, debug and simplify models, but can mislead if read as cause and effect.
- model-agnosticmodel-agnostic, adj.: Describes a method that works with any kind of AI model, because it only needs to feed the model inputs and observe its outputs, never to look inside.
Sources
- Apley, D. W. & Zhu, J. (2020). Visualizing the Effects of Predictor Variables in Black Box Supervised Learning Models. Journal of the Royal Statistical Society Series B 82(4), 1059–1086.
Mentioned in
global explanationglobal explanation, n.: An explanation of how an AI model behaves overall — which factors matter most across all its decisions, and in what way — rather than why it made one particular decision., partial dependence plotpartial dependence plot, n.: A chart showing how a model's average prediction changes as one input is varied while everything else stays as it is in the data — for example, predicted house price against floor area., individual conditional expectationindividual conditional expectation, n.: A chart that traces, for each example separately, what a model would predict as one input is swept across its range, exposing differences between examples that an overall average would hide., model-agnosticmodel-agnostic, adj.: Describes a method that works with any kind of AI model, because it only needs to feed the model inputs and observe its outputs, never to look inside.