lbfgsfloatval_t RAE::decay()
{
lbfgsfloatval_t val = 0;
for(int row = 0; row < weights1.rows(); row++)
{
for(int col = 0; col < weights1.cols(); col++)
{
val += pow(weights1(row, col), 2)/2;
val += pow(weights2(col, row), 2)/2;
}
}
for(int col = 0; col < weights_b1.cols(); col++)
{
val += pow(weights_b1(0, col), 2)/2;
}
for(int col = 0; col < weights_b2.cols(); col++)
{
val += pow(weights_b2(0, col), 2)/2;
}
return val;
}
GPU_PERF_TEST(BlendLinear, cv::gpu::DeviceInfo, cv::Size, MatType)
{
cv::gpu::DeviceInfo devInfo = GET_PARAM(0);
cv::gpu::setDevice(devInfo.deviceID());
cv::Size size = GET_PARAM(1);
int type = GET_PARAM(2);
cv::Mat img1_host(size, type);
fill(img1_host, 0, 255);
cv::Mat img2_host(size, type);
fill(img2_host, 0, 255);
cv::gpu::GpuMat img1(img1_host);
cv::gpu::GpuMat img2(img2_host);
cv::gpu::GpuMat weights1(size, CV_32FC1, cv::Scalar::all(0.5));
cv::gpu::GpuMat weights2(size, CV_32FC1, cv::Scalar::all(0.5));
cv::gpu::GpuMat dst;
cv::gpu::blendLinear(img1, img2, weights1, weights2, dst);
TEST_CYCLE()
{
cv::gpu::blendLinear(img1, img2, weights1, weights2, dst);
}
}
OCL_PERF_TEST_P(BlendLinearFixture, BlendLinear, ::testing::Combine(OCL_TEST_SIZES, OCL_TEST_TYPES_134))
{
Size_MatType_t params = GetParam();
const Size srcSize = get<0>(params);
const int srcType = get<1>(params);
const double eps = CV_MAT_DEPTH(srcType) <= CV_32S ? 1.0 : 0.2;
checkDeviceMaxMemoryAllocSize(srcSize, srcType);
UMat src1(srcSize, srcType), src2(srcSize, srcType), dst(srcSize, srcType);
UMat weights1(srcSize, CV_32FC1), weights2(srcSize, CV_32FC1);
declare.in(src1, src2, WARMUP_RNG).in(weights1, weights2, WARMUP_READ).out(dst);
randu(weights1, 0, 1);
randu(weights2, 0, 1);
OCL_TEST_CYCLE() cv::blendLinear(src1, src2, weights1, weights2, dst);
SANITY_CHECK(dst, eps);
}
/**
* score_categorical
*
* Perform scoring for a given precondition and variable.
*
* @param int r: Precondition in use
* @param long i: SNP in use
* @param Condition::comparison &: Type of comparison returned.
* @param short s: Optimal split point.
*/
double ADTree::score_categorical(int r, long i, Condition::comparison &c, short &s){
double *ret_score = new double[9];
double z1, z2, ret;
weights2(ret_score,r,i); // run both tests.
z2 = 2.0*(sqrt(ret_score[0]*ret_score[1]) + sqrt(ret_score[2]*ret_score[3])) + ret_score[8];
z1 = 2.0*(sqrt(ret_score[4]*ret_score[5]) + sqrt(ret_score[6]*ret_score[7])) + ret_score[8];
if(z1 < z2){
ret = z1;
c = Condition::GE;
s = 4;
}else{ // so default to z2.
ret = z2;
c = Condition::GE;
s = 2;
}
delete[] ret_score;
return ret;
}
void RAE::loadWeights(Parameter* para)
{
string filename;
if(RAEType == SL)
{
filename = para->getPara("srcRAEWeightsLogFile");
}
else
{
filename = para->getPara("tgtRAEWeightsLogFile");
}
ifstream in(filename.c_str(), ios::in);
bool rae_w1 = false;
bool rae_b1 = false;
bool rae_w2 = false;
bool rae_b2 = false;
int row = 0;
string line;
while(getline(in, line))
{
if(line.find("W1") == 0)
{
row = 0;
rae_w1 = true;
continue;
}
if(line.find("W2") == 0)
{
row = 0;
rae_b1 = false;
rae_w2 = true;
continue;
}
if(line.find("B1") == 0)
{
row = 0;
rae_w1 = false;
rae_b1 = true;
continue;
}
if(line.find("B2") == 0)
{
row = 0;
rae_w2 = false;
rae_b2 = true;
continue;
}
if(rae_w1)
{
stringstream ss(line);
for(int col = 0; col < weights1.cols(); col++)
{
ss >> weights1(row, col);
}
}
if(rae_w2)
{
stringstream ss(line);
for(int col = 0; col < weights2.cols(); col++)
{
ss >> weights2(row, col);
}
}
if(rae_b2)
{
stringstream ss(line);
for(int col = 0; col < weights_b2.cols(); col++)
{
ss >> weights_b2(row, col);
}
}
if(rae_b1)
{
stringstream ss(line);
for(int col = 0; col < weights_b1.cols(); col++)
{
ss >> weights_b1(row, col);
}
}
row++;
}
}