Overview¶
- Data Acquisition
- Learn
- Estimate
- Predict
Data¶
D={xn,yn,zn∗,}n=1N Measurements:Covariates:Simulationed States:Reanalysis States:ynxnznsimzn∗∈Y⊆RDy∈X⊆RDx∈Zsim⊆RDz∈Z∗⊆RDz
Learning¶
I have data, D, which captures the phenomena that I want to learn.
I want to learn a model, f, with the associated parameters, θ, give the data, D.
θ∗=θargminL(θ;D) where L(⋅) is our loss function.
L:RDθ×D→R
Estimation¶
I have a model, f, and parameters, θ.
I have some measurements, y.
I want to estimate a state, z.
z∗(θ)=zargminJ(z;θ,D) where J(⋅) is our objective function defined as:
J:RDz×RDθ×D→R
Parameter & State Estimation¶
Parameter Estimation:State Estimation:θ∗=θargminL(θ;D)z∗(θ)=zargminJ(z;θ,D) This is akin to the:
- Approximate Inference methods - expectaction maximization, variational inference
- Bi-Level Optimization
- Data Assimilation
Prediction¶
I have my model, parameters, and state estimation.
I want to make a prediction for my QoI, u.
u∗=f(z∗,θ) In this case, we never have access to any sort of validation, u.
We are simply making a prediction.