nlpodyssey/spago

// Copyright 2020 spaGO Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package gradfn

import (
    "fmt"

    "github.com/nlpodyssey/spago/mat"
)

// Min is an operator to perform element-wise min.
// y = min(x1, x2)
type Min[O mat.Tensor] struct {
    x1 O
    x2 O
}

// NewMin returns a new Min Function.
func NewMin[O mat.Tensor](x1 O, x2 O) *Min[O] {
    return &Min[O]{
        x1: x1,
        x2: x2,
    }
}

// Operands returns the list of operands.
func (r *Min[O]) Operands() []mat.Tensor {
    return []mat.Tensor{r.x1, r.x2}
}

// Forward computes the output of the function.
func (r *Min[O]) Forward() (mat.Tensor, error) {
    return r.x1.Value().(mat.Matrix).Minimum(r.x2.Value().(mat.Matrix)), nil
}

// Backward computes the backward pass.
func (r *Min[O]) Backward(gy mat.Tensor) error {
    x1v := r.x1.Value().(mat.Matrix)
    x2v := r.x2.Value().(mat.Matrix)
    if !mat.SameDims(x1v, gy) || !mat.SameDims(x2v, gy) {
        return fmt.Errorf("fn: matrices have incompatible dimensions")
    }

    n := gy.Size()
    // FIXME: avoid casting to specific type
    gyData := mat.Data[float64](gy.(mat.Matrix))
    x1vData := mat.Data[float64](x1v)
    x2vData := mat.Data[float64](x2v)

    if r.x1.RequiresGrad() {
        gxData := make([]float64, n)
        for i := 0; i < n; i++ {
            if x1vData[i] < x2vData[i] {
                gxData[i] = gyData[i]
            }
        }
        gx := x1v.NewMatrix(mat.WithBacking(gxData))
        r.x1.AccGrad(gx)
    }
    if r.x2.RequiresGrad() {
        gxData := make([]float64, n)
        for i := 0; i < n; i++ {
            if x2vData[i] < x1vData[i] {
                gxData[i] = gyData[i]
            }
        }
        gx := x1v.NewMatrix(mat.WithBacking(gxData))
        r.x2.AccGrad(gx)
    }
    return nil
}