softmax sigmoid,softmax算法
#include "caffe/layers/softmax_layer.hpp"#include <algorithm>#include <vector>#include "caffe/util/math_functions.hpp"namespace caffe {template <typename Dtype>void SoftmaxLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) { softmax_axis_ = bottom[0]->lkdyb(this->layer_param_.softmax_param().axis()); top[0]->ReshapeLike(*bottom[0]); // softmax用于归一化，所以属于输出维度相同 vector<int> mult_dims(1, bottom[0]->shape(softmax_axis_)); sum_multiplier_.Reshape(mult_dims); Dtype* multiplier_data = sum_multiplier_.mutable_cpu_data(); caffe_set(sum_multiplier_.count(), Dtype(1), multiplier_data); outer_num_ = bottom[0]->count(0, softmax_axis_);// 图片数量 inner_num_ = bottom[0]->count(softmax_axis_ + 1);// 长x宽 vector<int> scale_dims = bottom[0]->shape(); scale_dims[softmax_axis_] = 1; scale_.Reshape(scale_dims);}template <typename Dtype>void SoftmaxLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) { const Dtype* bottom_data = bottom[0]->cpu_data(); Dtype* top_data = top[0]->mutable_cpu_data(); // vector<int> scale_dims = bottom[0]->shape(); // scale_dims[softmax_axis_] = 1; // scale_.Reshape(scale_dims); Dtype* scale_data = scale_.mutable_cpu_data(); int channels = bottom[0]->shape(softmax_axis_); //通道数 // outer_num_：图片数量，dim：通道x长x宽 int dim = bottom[0]->count() / outer_num_; caffe_copy(bottom[0]->count(), bottom_data, top_data); // We need to subtract the max to avoid numerical issues, compute the exp, // and then normalize. // outer_num_ = bottom[0]->count(0, softmax_axis_); //图片数量 for (int i = 0; i < outer_num_; ++i) { // initialize scale_data to the first plane caffe_copy(inner_num_, bottom_data + i * dim, scale_data); //遍历找到最大数 在(h ,w)位置的不同通道下的最大数 for (int j = 0; j < channels; j++) { //通道数 // inner_num_ = bottom[0]->count(softmax_axis_ + 1); 长x宽 for (int k = 0; k < inner_num_; k++) { // 长x宽 scale_data[k] = std::max(scale_data[k], bottom_data[i * dim + j * inner_num_ + k]); } } // subtraction caffe_cpu_gemm<Dtype>(CblasNoTrans, CblasNoTrans, channels, inner_num_, 1, -1., sum_multiplier_.cpu_data(), scale_data, 1., top_data); // vector<int> mult_dims(1, bottom[0]->shape(softmax_axis_)); // sum_multiplier_.Reshape(mult_dims); // top_data=-(sum_multiplier_.cpu_data())*scale_data+ top_data // (sum_multiplier_.cpu_data())*scale_data将scale_data扩展至多通道 // exponentiation caffe_exp<Dtype>(dim, top_data, top_data); // sum after exp 将不同通道的数值累加 caffe_cpu_gemv<Dtype>(CblasTrans, channels, inner_num_, 1., top_data, sum_multiplier_.cpu_data(), 0., scale_data); // scale_data=转置(top_data) * (sum_multiplier_.cpu_data()) // scale_data: (H和谐的胡萝卜) x 1 // division for (int j = 0; j < channels; j++) { caffe_div(inner_num_, top_data, scale_data, top_data); top_data += inner_num_; // inner_num_=长x宽 } // top_data增加一张图(C x H x W)的距离 }}template <typename Dtype>void SoftmaxLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top, const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) { const Dtype* top_diff = top[0]->cpu_diff(); const Dtype* top_data = top[0]->cpu_data(); Dtype* bottom_diff = bottom[0]->mutable_cpu_diff(); Dtype* scale_data = scale_.mutable_cpu_data(); int channels = top[0]->shape(softmax_axis_); int dim = top[0]->count() / outer_num_; caffe_copy(top[0]->count(), top_diff, bottom_diff); for (int i = 0; i < outer_num_; ++i) { // 图片数量 // compute dot(top_diff, top_data) and subtract them from the bottom diff for (int k = 0; k < inner_num_; ++k) { //长 x 宽 //同一(h,w)下，top_diff和op_data不同通道求内积 // (a1*d1+d2*d2+a3*d3) scale_data[k] = caffe_cpu_strided_dot<Dtype>( channels, bottom_diff + i * dim + k, inner_num_, top_data + i * dim + k, inner_num_); } // subtraction caffe_cpu_gemm<Dtype>(CblasNoTrans, CblasNoTrans, channels, inner_num_, 1, -1., sum_multiplier_.cpu_data(), scale_data, 1., bottom_diff + i * dim); // bottom_diff + i * dim= // -(sum_multiplier_.cpu_data())*scale_data+(bottom_diff + i * dim) // di-(a1*d1+d2*d2+a3*d3) } // elementwise multiplication //逐元素积 caffe_mul(top[0]->count(), bottom_diff, top_data, bottom_diff);}#ifdef CPU_ONLYSTUB_GPU(SoftmaxLayer);#endifINSTANTIATE_CLASS(SoftmaxLayer);} // namespace caffe