Softmax annotated: Difference between revisions
From Algolit
| Line 8: | Line 8: | ||
|} | |} | ||
| − | '''''Softmax annotated''''' is a annotated script of a softmax function written in python. The basis of the code is found at the [https://en.wikipedia.org/wiki/Softmax_function#Artificial_neural_networks Softmax function] page on the English Wikipedia. This script was part of proces to get a better understanding of the mathematics behind neural networks. The softmax function is | + | '''''Softmax annotated''''' is a annotated script of a softmax function written in python. The basis of the code is found at the [https://en.wikipedia.org/wiki/Softmax_function#Artificial_neural_networks Softmax function] page on the English Wikipedia. This script was part of proces to get a better understanding of the mathematics behind neural networks. |
| + | |||
| + | The softmax function provides a very basic classifier based on logistic regression. It is a building block used in neural networks. A neural network can be seen as running multiple logistic regression in parallel, which then feed into another layer of (multiple) logistic regressions and so on. | ||
| + | |||
| + | More information see [https://cs224d.stanford.edu/lectures/CS224d-Lecture4.pdf Lecture 4 of the Stanford Deep Learning course] | ||
Revision as of 17:50, 25 October 2017
| Type: | Algolit extension |
| Technique(s): | mathematics, softmax function, python |
| Developed by: | Algolit |
Softmax annotated is a annotated script of a softmax function written in python. The basis of the code is found at the Softmax function page on the English Wikipedia. This script was part of proces to get a better understanding of the mathematics behind neural networks.
The softmax function provides a very basic classifier based on logistic regression. It is a building block used in neural networks. A neural network can be seen as running multiple logistic regression in parallel, which then feed into another layer of (multiple) logistic regressions and so on.
More information see Lecture 4 of the Stanford Deep Learning course