My question regards to the Convolutional Neural Networks (CNNs)
What is the relationship between Non-trainable parameters vs. the size of the data set?
How we can make the Non-trainable parameters constant?
It can be calculated by considering computing specifications and the number of trainable parameters.
Have a look at :: https://www.iditect.com/how-to/51729624.html
https://stackoverflow.com/questions/47312219/what-is-the-definition-of-a-non-trainable-parameter
https://www.reddit.com/r/deeplearning/comments/9gtc9s/trainable_vs_non_trainable_parameters/
Dear Fadil Al-Jaberi ,
Look the link, maybe useful.
https://stackoverflow.com/questions/50144597/if-we-combine-one-trainable-parameters-with-a-non-trainable-parameter-is-the-or/50215456
Regards,
Shafagat
You can calculate the number of trainable parameters and memory requirements for a CNN based on its non-trainable parameters such as convolutinal filter sizes, stride and so on. The attached paper provides means of calculations as well as the results for CNNs in image recognition based on popular benchmarks from 2012 to 2017:
Conference Paper Is Bigger Always Better? Lessons Learnt from the Evolution o...
see
https://www.reddit.com/r/deeplearning/comments/9gtc9s/trainable_vs_non_trainable_parameters/
It can be calculated by considering computing specifications and the number of trainable parameters.
- Badges
- Science topic