error when run the test example

[XMUSpeechLab]

Hello, I run the following example code, but got error for the umappp::initialize function.

#include "umappp/umappp.hpp"
#include "umappp/knncolle/knncolle.hpp"
#include
#include
#include

int main() {
// 1. 准备高维数据（示例数据）
const int ndim = 5; // 高维数据维度
const int nobs = 1000; // 样本数量
std::vector data(ndim * nobs);

// 生成随机高维数据（实际应用中替换为你的数据）
std::mt19937_64 rng(42);
std::normal_distribution<double> dist(0, 1.0);
for (int i = 0; i < ndim * nobs; ++i) {
    data[i] = dist(rng);
}

// 2. 配置UMAP参数
umappp::Options opt;
opt.num_neighbors = 15;      // 近邻数量
opt.num_epochs = 500;        // 迭代次数
opt.min_dist = 0.1;          // 最小距离
opt.learning_rate = 1.0;     // 学习率
opt.num_threads = 4;         // 线程数

// 3. 配置近邻搜索算法（使用VP树+欧氏距离）
knncolle::VptreeBuilder<int, double, double> vp_builder(
    std::make_shared<knncolle::EuclideanDistance<double, double>>()
);

// 4. 初始化嵌入结果存储
const int out_dim = 2;       // 降维到2维（可视化常用）
std::vector<double> embedding(nobs * out_dim);

// 5. 初始化UMAP状态
auto status = umappp::initialize(
    ndim,
    nobs,
    data.data(),
    vp_builder,
    out_dim,
    embedding.data(),
    opt
);

// 6. 运行UMAP降维算法
status.run(embedding.data());

// 7. 输出结果（前5个样本的2维坐标）
std::cout << "UMAP降维结果（前5个样本）：" << std::endl;
for (int i = 0; i < 5; ++i) {
    std::cout << "样本 " << i << ": ("
        << embedding[i * out_dim] << ", "
        << embedding[i * out_dim + 1] << ")" << std::endl;
}

return 0;

}

further debug to the error point, it locate at the following part of GenealBlcokPanelKernel.h in Eigen:
// copy the remaining columns one at a time (nr==1)
for (Index j2 = packet_cols4; j2 < cols; ++j2) {
if (PanelMode) count += offset;
const LinearMapper dm0 = rhs.getLinearMapper(0, j2);
for (Index k = 0; k < depth; k++) {
blockB[count] = cj(dm0(k));
count += 1;
}
if (PanelMode) count += (stride - offset - depth);
}

Can anyone give some tips or help? Thanks!

[LTLA]

I don't see any errors in the README example besides a few misnamed variables (since corrected). My code is:

#include "umappp/umappp.hpp"

int main(){
// Assuming `data` contains high-dimensional data in column-major format,
// i.e., each column is a observation and each row is a dimension.
// We fill it with some random noise for the purposes of this example.
int ndim = 10;
int nobs = 2000;
std::vector<double> data(ndim * nobs);
std::mt19937_64 rng(1000);
std::normal_distribution<double> ndist;
for (auto& x : data) {
    x = ndist(rng);
}

// Configuring the neighbor search algorithm; here, we'll be using an exact
// search based on VP trees with a Euclidean distance metric.
knncolle::VptreeBuilder<int, double, double> vp_builder(
    std::make_shared<knncolle::EuclideanDistance<double, double> >()
);

// Set number of dimensions in the output embedding.
size_t out_dim = 2;
std::vector<double> embedding(nobs * out_dim);

// Initialize the UMAP state:
umappp::Options opt;
auto status = umappp::initialize(
    ndim,
    nobs,
    data.data(),
    vp_builder, 
    out_dim,
    embedding.data(),
    opt
);

// Run UMAP algorithm to completion. This updates the contents
// of the 'embedding' vector supplied to initialize().
status.run(embedding.data());
return 0;
}

Compiled in a subdirectory of this repository, with:

g++ --std=c++17 \
    -I../include \
    -I../build/_deps/aarand-src/include \
    -I../build/_deps/eigen-src \
    -I../build/_deps/knncolle-src/include \
    -I../build/_deps/sanisizer-src/include \
    -I../build/_deps/subpar-src/include \
    -I../build/_deps/irlba-src/include \
    test.cpp 

And it seems to run to completion:

$ time ./a.out 
real	0m4.111s
user	0m4.111s
sys	0m0.000s

$ echo $?
0