A new test for comparing two sets of high-dimensional data has been developed. This test is able to adapt to different shapes of data distributions and can accurately detect differences between the two sets. The test performs well in simulations and real-world applications, outperforming other existing methods.