\newcommand{\par}[1] {\left({#1}\right)} \newcommand{\arr}[1] {\left[{#1}\right]} \newcommand{\hsh}[1] {\left\{{#1}\right\}} \newcommand{\abs}[1] {\left|{#1}\right|} \newcommand{\basisCanon} {\mathbf{\hat{e}}} \newcommand{\basisCanonElement}[1][i] {\basisCanon_{#1}} \newcommand{\varietyAffine}[2][X] {\mathscr{#1}\par{#2}} \newcommand{\smol} {\epsilon} \newcommand{\derivativeDirectionalSimple}[3][f] {{#1}'\par{#2_0; #3}} \newcommand{\derivativeDirectional}[3][f] {\frac{d #1 \par{#2_0; #3}}{d \smol}} \newcommand{\derivativePartial}[3][f] {\frac{\partial #1 \par{#2_0}}{\partial x_{#3}}} \newcommand{\derivativePartialOutput}[4][f] {\frac{\partial #1_{#3} \par{#2_0}}{\partial x_{#4}}} \newcommand{\derivativeGradientSimple}[2][f] {\nabla_{#1} \par{#2_0}} \newcommand{\continueField}[1][1] {\mathcal{C}^{#1}} \newcommand{\hessianSimple}[2][f] {\mathbf{\nabla}^2_{#1} \par{#2_0}} \newcommand{\hessianElement}[4][f] {\frac{\partial^2 #1 \par{#2}}{\partial #2_{#3}\ \cdot\ \partial #2_{#4}}} \newcommand{\iacobianSimple}[2][f] {\mathbf{J}_{#1} \par{#2}} \newcommand{\fmlDataset} {\mathbf{D}} \newcommand{\fmlDatasetSize} {P} \newcommand{\fmlDatasetTrain} {\fmlDataset_{train}} \newcommand{\fmlDatasetValidate} {\fmlDataset_{validate}} \newcommand{\fmlDatasetTest} {\fmlDataset_{test}} \newcommand{\fmlInputs} {\mathbf{X}} \newcommand{\fmlIndex} {i} \newcommand{\fmlInput}[1][\fmlIndex] {\mathbf{x}_{#1}} \newcommand{\fmlInputSize} {n} \newcommand{\fmlInputField}[1][\fmlInputSize] {\mathbb{X}^{#1}} \newcommand{\fmlElementIndex} {j} \newcommand{\fmlInputElement}[1][\fmlIndex\fmlElementIndex] {x_{#1}} \newcommand{\fmlInputElementField} {\mathbb{X}} \newcommand{\fmlOutputs} {\mathbf{y}} \newcommand{\fmlOutput}[1][\fmlIndex] {y_{#1}} \newcommand{\fmlOutputField} {\mathbb{Y}} \newcommand{\fmlClassifier} {f} \newcommand{\fmlModel} {\fmlClassifier^*} \newcommand{\fmlHypothesisSpace} {\mathcal{H}} \newcommand{\fmlComplexity} {c} \newcommand{\fmlComplexityCoefficient} {\lambda} \newcommand{\fmlLoss} {V}