reductions.hpp File Reference

Defines reduction operations for tensor expressions. More...

#include <array>
#include <cassert>
#include "concepts.hpp"
#include "types.hpp"
#include "traits.hpp"
#include "shape.hpp"
#include "tensor.hpp"
#include "indexing.hpp"
#include "exceptions.hpp"

Include dependency graph for reductions.hpp:

Go to the source code of this file.

Classes
class	tannic::expression::Reduction< Reducer, Operand >
	Lazy reduction expression. More...
struct	tannic::expression::Argmax
	Finds the indices of maximum values along an axis. More...
struct	tannic::expression::Argmin
	Finds the indexes of minimum values along an axis. More...
struct	tannic::expression::Argsum
	Sums tensor values along an axis. More...
struct	tannic::expression::Argmean
	Computes the mean along an axis. More...

Namespaces
namespace	tannic
namespace	tannic::expression

Functions
template<Expression Source>
constexpr auto	tannic::expression::argmax (Source &&source, int axis, bool keepdim=false)
	Creates an Argmax reduction.
template<Expression Source>
constexpr auto	tannic::expression::argmin (Source &&source, int axis, bool keepdim=false)
	Creates an Argmin reduction.
template<Expression Source>
constexpr auto	tannic::expression::sum (Source &&source, int axis, bool keepdim=false)
	Creates a sum reduction.
template<Expression Source>
constexpr auto	tannic::expression::mean (Source &&source, int axis, bool keepdim=false)
	Creates a mean reduction.

Detailed Description

Defines reduction operations for tensor expressions.

Author

Eric Hermosis

Date

2025

This header provides reduction operations for tensor-like objects, implemented as expression templates. Currently supports:

• argmax: Index of maximum value along axis optionally keeping the original rank.
• argmin: Index of minimum value along axis optionally keeping the original rank.
• sum: Sum of the values along axis optionally keeping the original rank.
• mean: Mean of the values along axis optionally keeping the original rank.

All reductions are lazy-evaluated and maintain proper shape/dtype transformations. Part of the Tannic Tensor Library.