We present a novel approach for gaussian process inference, by formulating kernel inverses as approximate solutions of linear equations using convolutions and optimize the model with conjugate gradient descent.
Tensorflow, PyTorch and GPytorch
Please see the final report: https://github.com/AnuroopKuppam/gp-inference-with-convolutions/blob/master/write%20ups/final.pdf