Derivative Hadamard Product
Answer
Answers can only be viewed under the following conditions:
- The questioner was satisfied with and accepted the answer, or
- The answer was evaluated as being 100% correct by the judge.
1 Attachment
Mathe
3.5K
The answer is accepted.
Join Matchmaticians Affiliate Marketing
Program to earn up to a 50% commission on every question that your affiliated users ask or answer.
- answered
- 1242 views
- $58.00
Related Questions
- Suppose that $T \in L(V,W)$. Prove that if Img$(T)$ is dense in $W$ then $T^*$ is one-to-one.
- Problem Help - Matrix/Markov
- The Span and Uniqueness of Solutions in a Parametric Matrix
- Prove Property of Projection Matrices
- Find eigenvalues and eigenvectors of $\begin{pmatrix} 1 & 6 & 0 \\ 0& 2 & 1 \\ 0 & 1 & 2 \end{pmatrix} $
- Use Rouche’s Theorem to show that all roots of $z ^6 + (1 + i)z + 1 = 0$ lines inside the annulus $ \frac{1}{2} \leq |z| \leq \frac{5}{4}$
- Elementary row reduction for an $n\times n$ matrix
- Exercise 4.33 from Spivak's Calculus on Manifolds.
What do you mean with the derivative wrt W?
I did d some research and it seems to me that there is no commonly accepted definition. Please provide one.
The term is part of the backpropagation of my neural network. I need to calculate this derivative to update my weights in the neural network. Hope this helps.
The hadamard product is the elemt-wise product of two matrices of the same dimensionality https://en.m.wikipedia.org/wiki/Hadamard_product_(matrices)
One can view the product as a function from R^(|V|\times |d|) to R^(|d|). So the derivative may mean just the matrix of partial derivatives. Do you expect the result to be a a |d| by ( |V|\times |d| ) matrix?
a and b must be row vectors for this product to make sense.
Ideally the result would be in the dimenson of |V|
The preferred notation for a vector is as a column vector, so ideally aW would be written as a^t W , meaning a transpose times W. will keep with the notation you used, so that aW is a row vector times a matrix, but this is not a very common notation.
I will keep with the notation you used, so that aW is a row vector times a matrix, but this is not a very common notation*
I am sorry for any inconvenience with the notation. Of course, a and b would be transposed.