1,前向传播
分为两步：1，计算softmax概率prob_data，直接使用softmaxlayer的forward函数；
2，计算loss，采用交叉熵，即每个第i类数据的loss为-log(prob(i))。
template <typename Dtype> void SoftmaxWithLossLayer<Dtype>::Forward_cpu( const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) { // The forward pass computes the softmax prob values. softmax_layer_->Forward(softmax_bottom_vec_, softmax_top_vec_);//直接使用softmax_layer->forward() const Dtype* prob_data = prob_.cpu_data(); //概率数据 const Dtype* label = bottom[1]->cpu_data(); //真实标签 int dim = prob_.count() / outer_num_; int count = 0; Dtype loss = 0; for (int i = 0; i < outer_num_; ++i) { for (int j = 0; j < inner_num_; j++) { const int label_value = static_cast<int>(label[i * inner_num_ + j]); if (has_ignore_label_ && label_value == ignore_label_) { continue; } DCHECK_GE(label_value, 0); DCHECK_LT(label_value, prob_.shape(softmax_axis_)); loss -= log(std::max(prob_data[i * dim + label_value * inner_num_ + j], Dtype(FLT_MIN))); //每个数据i的损失为-log(prob(i)) ++count; } } //loss除去样本总数，得到每个样本的loss top[0]->mutable_cpu_data()[0] = loss / get_normalizer(normalization_, count); if (top.size() == 2) { top[1]->ShareData(prob_); } }2,反向传播
template <typename Dtype> void SoftmaxWithLossLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top, const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) { if (propagate_down[1]) { LOG(FATAL) << this->type() << " Layer cannot backpropagate to label inputs."; } if (propagate_down[0]) { Dtype* bottom_diff = bottom[0]->mutable_cpu_diff(); const Dtype* prob_data = prob_.cpu_data(); caffe_copy(prob_.count(), prob_data, bottom_diff);//把概率数据复制到bottom_diff const Dtype* label = bottom[1]->cpu_data(); //获得标签数据 int dim = prob_.count() / outer_num_; int count = 0; for (int i = 0; i < outer_num_; ++i) { for (int j = 0; j < inner_num_; ++j) { const int label_value = static_cast<int>(label[i * inner_num_ + j]);//获得真实标签 if (has_ignore_label_ && label_value == ignore_label_) { //如果忽略标签，则偏导数为0 for (int c = 0; c < bottom[0]->shape(softmax_axis_); ++c) { bottom_diff[i * dim + c * inner_num_ + j] = 0; } } else { //计算当前概率密度与理想概率密度之差（label位对应的理想概率密度为1，其他为0，故不计算） bottom_diff[i * dim + label_value * inner_num_ + j] -= 1; ++count; } } } // Scale gradient //缩放 Dtype loss_weight = top[0]->cpu_diff()[0] / get_normalizer(normalization_, count); caffe_scal(prob_.count(), loss_weight, bottom_diff); } }
