3 ms·
>> CNNs use convolutions which are a generalization of matrix multiplication. > A nitpick: CNNs are most definitely not a generalization of matrix multiplicati
by dkislyuk 6y ago
>> CNNs use convolutions which are a generalization of matrix multiplication.
> A nitpick: CNNs are most definitely not a generalization of matrix multiplication. In fact, the opposite: you can view CNNs as a matrix multiplication with a particular matrix structure.
Neither is really a generalization. Any matrix multiplication can be implemented with a convolution function, and yet any convolution can be represented as a matrix multiplication via im2col.
- bnegreve 6y agoDiscrete convolutions are a special case of matrix multiplication in the sense that a discrete convolution can be formulated as the product between a Toeplitz matrix and the input signal. [1] https://en.wikipedia.org/wiki/Toeplitz_matrix#Discrete_convolution https://en.wikipedia.org/wiki/Toeplitz_matrix#Discrete_convo...