How does tf.einsum() in tensorflow calculates gradients for matrix multiplication?
By : Nikhil Chaurasia
Date : March 29 2020, 07:55 AM
wish help you to fix your issue I disagree with strangepoop's answer, mostly in the idea that "If you understand how backprop works in those, you can desugar to understand backprop here." einsum is an elegant operation that is more fundamental than matmul or any other tensor operation. Understanding backpropagation in matmul only amounts to understanding a special case of einsum and presents a very limited view.