Yunus Cobanoglu

but what about computing the Monte Carlo Estimate for Infinite Width so $\mathbb{E}(softmax(u)softmax(v))$ with $u,v$ from a multivariate Gaussian with a block symmetric covariance ?

ok thanks there is also [Exploring Alternatives to Softmax Function](https://arxiv.org/pdf/2011.11538.pdf) which contains a Taylor Approximation of Softmax maybe there is a closed form solution for the taylor approximation. I would...

I have done work on [Graph Neural Gaussian Processes and Graph Neural Tangent Kernel for Node Classification/ Regression](https://github.com/yCobanoglu/infinite-width-gnns) (transductive learning) using the Neural Tangent Library.

I couldn't get it to work with custom dataset.

Check [my repo](https://github.com/yCobanoglu/infinite-width-gnns) on transductive Node Classification/ Regression using Graph Neural Network Gaussian Processes and Graph Neural Tangent Kernel using the Neural Tangents Library.

I could fix it by installing [mongo](https://wiki.archlinux.org/title/MongoDB) for ARCH. You can close the issue.