adapters.hpp File Reference

Contains some general purpose loss adapters. More...

#include <concepts>

Include dependency graph for adapters.hpp:

Go to the source code of this file.

Functions
auto	dl::lossAdapter (auto lossObjective)
	A training adapter for applying a single loss objective.

Detailed Description

Contains some general purpose loss adapters.

Definition in file adapters.hpp.

Function Documentation

lossAdapter()

auto dl::lossAdapter

(

auto

lossObjective

)

A training adapter for applying a single loss objective.

Let \(\mathcal M\colon X\to Y\) denote the model to be trained or evaluated, where \(X\) is the space of all inputs and \(Y\) is the space of all outputs. Further, let \(\mathcal L\colon Y\times Y \to \mathbb R\) denote a loss objective. Given the input \(x\in X\) and desired output \(y\in Y\), the loss adapter returns

\[\mathcal{L}(\mathcal{M}(x), y).\]

Or, in short: The loss adapter tells the trainer to optimize the given loss objective.

Parameters

lossObjective

the loss objective to be optimized for.

Returns

A loss adapter for optimizing the specified loss objective.

Definition at line 25 of file adapters.hpp.

                                         {
        constexpr auto func = [](auto loss, auto& model, auto& x, auto& y) { return loss(model(x), y); };
        return std::bind_front(func, lossObjective);
    }