Currently i have come across adjoint method for optimization problems; and i have found that this method can be used as an alternate for chain rule. This made me confused. If anyone explains the mechanism, with example, of adjoint method for optimization in place of chain rule, I would be very thankful!
Asked
Active
Viewed 164 times
0
-
1Welcome to math SE. Do you have an example in mind? Have you tried anything? – Alain Remillard Jan 16 '20 at 18:08
-
Look up reverse mode of automatic differntiation, also known in neural networks as backpropagation. – Mark L. Stone Jan 16 '20 at 18:12
-
@AlainRemillard, thank you for the response. Actually i have come across this while reading a paper related to this, where the author has used adjoint method in place of chain rule for finding gradients of loss function w.r.t to weights. The approach is explained here link – NeuralAI Jan 16 '20 at 18:15
-
yes @MarkL.Stone! I have read it there too! I just want to know about it: any general formula or method explanation in general. – NeuralAI Jan 16 '20 at 18:18
-
I am telling you to read material on reverse mode of automatic differentiation or the use of backpropagation to compute gradients in neural networks. – Mark L. Stone Jan 16 '20 at 18:20