Konstantin Eckle

Postdoctoral Researcher

Leiden University

I am a postdoctoral researcher in statistics at the mathematical institute of Leiden University. My research focuses on statistical theory for deep learning and multiscale testing in direct and inverse problems.

I work in collarboration with Johannes Schmidt-Hieber on L2 risk bounds for deep feedforward and convolutional neural networks with ReLU activation function. We consider regression and classification frameworks.

In multiscale testing many local tests for different locations and bandwidths are performed simultaneously. We use recent high-dimensional Gaussian approximation results to derive quantiles which control the family-wise error rate of the testing procedure. My interest lies in tests for shape constraints of a density and I focus now on inverse problems in econometrics in collaboration with Fabian Dunker.

I am also a proud member of Mathletico, the soccer team of the mathematical institute.