partial dependence plot
In plain wordsA 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.
A 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. plot of a model's partial dependence on one or two features: the prediction averaged over the data distribution of the remaining features, as a function of the chosen ones. It is estimated by setting the chosen feature to each value on a grid for every instance in a dataset, predicting, and averaging. Because the average is taken over the marginal distribution of the other features, the plot reveals the shape of a feature's effect (linear, threshold, saturating) for any black-boxmodel-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. model and, with two features, their interaction. It assumes the chosen feature is roughly independent of the others: under correlation the grid creates unrealistic combinations (a 20-square-meter house with eight bedrooms) and the curve can mislead, a weakness addressed by accumulated local effectsaccumulated local effects, n.: A 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.. Averaging can also hide opposite effects in subgroups, which 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. curves reveal.
The partial dependence plot showed predicted default risk flattening once income passed about $80,000.
Origin
Introduced by Jerome H. Friedman in 'Greedy Function Approximation: A Gradient Boosting Machine' (Annals of Statistics, 2001), the paper that formalized gradient boostinggradient boosting, n.: A method that builds a strong predictor by adding small decision trees one at a time, each trained to correct the errors that remain after all the trees before it., as a way to interpret the resulting models.
Usage note
A partial dependence curve describes the model, not the world; it is not a causal effect unless strong assumptions hold.
Related
- 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.
- accumulated local effectsaccumulated local effects, n.: A 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.
- 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.
- gradient boostinggradient boosting, n.: A method that builds a strong predictor by adding small decision trees one at a time, each trained to correct the errors that remain after all the trees before it.
- SHAPSHAP, abbr.: A popular method that explains a single prediction by fairly dividing it among the inputs, using a rule from game theory for splitting a team's winnings among its players.
- 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.
Sources
- Friedman, J. H. (2001). Greedy Function Approximation: A Gradient Boosting Machine. Annals of Statistics 29(5), 1189–1232.
Mentioned in
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., generalized additive modelgeneralized additive model, n.: A prediction model that adds up separate, possibly curved effects of each input, so each input's influence can be drawn as its own graph and read off directly., 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., post hoc explanationpost hoc explanation, n.: An explanation produced after a model has been built, by a separate tool that examines the finished model, rather than one that comes from a model designed to be understandable., SHAPSHAP, abbr.: A popular method that explains a single prediction by fairly dividing it among the inputs, using a rule from game theory for splitting a team's winnings among its players., accumulated local effectsaccumulated local effects, n.: A 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., 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.