Predictors
The main entry point for embedding prediction models into JuMP is add_predictor.
All methods use the form y, formulation = MathOptAI.add_predictor(model, predictor, x) to add the relationship y = predictor(x) to model.
Supported predictors
The following predictors are supported. See their docstrings for details:
| Predictor | Problem class | JuMP | ExaModels |
|---|---|---|---|
Affine | Linear | Yes | Yes |
AffineCombination | Linear | Yes | |
AvgPool2d | Linear | Yes | |
BinaryDecisionTree | Mixed-integer linear | Yes | |
Conv2d. | Linear | Yes | |
GCNConv | Linear | Yes | |
GELU | Global nonlinear | Yes | Yes |
GrayBox | Local nonlinear | Yes | |
LayerNorm | Global nonlinear | Yes | |
LeakyReLU | Depends on inner ReLU | Yes | Yes |
MaxPool2d | Global nonlinear | Yes | |
MaxPool2dBigM | Mixed-integer linear | Yes | |
Pipeline | Yes | Yes | |
Quantile | Local nonlinear | Yes | |
ReLU | Global nonlinear | Yes | Yes |
ReLUBigM | Mixed-integer linear | Yes | |
ReLUQuadratic | Non-convex quadratic | Yes | |
ReLUSOS1 | Mixed-integer linear | Yes | |
Scale | Linear | Yes | Yes |
Sigmoid | Global nonlinear | Yes | Yes |
SoftMax | Global nonlinear | Yes | Yes |
SoftPlus | Global nonlinear | Yes | Yes |
TAGConv | Linear | Yes | |
Tanh | Global nonlinear | Yes | Yes |