Neural Networks on Going the distance
https://jeiwan.net/tags/neural-networks/
Recent content in Neural Networks on Going the distanceHugo -- gohugo.ioenSat, 05 Aug 2017 20:34:21 +0700TIL: Convolutional Filters Are Weights
https://jeiwan.net/posts/til-convolution-filters-are-weights/
Sat, 05 Aug 2017 20:34:21 +0700https://jeiwan.net/posts/til-convolution-filters-are-weights/It’s common knowledge that there’re weights in a fully-connected neural network. And these weights are not constant and are adjusted by an optimization algorithm (like gradient descent). Moreover, training a neural network actually means finding proper weights.
When I started learning about convolution and convolutional networks, first thing I was introduced to is filters. Filters are said to be matrices that are applied to an image to distort it in a certain way, unveiling certain aspect of an image.