How the processing is going on computaing neurons. why because in traditional machine learning algorithms we have only one weigth for each feature. But in deep learning for each feature we have different weights.
Hi @RajendraAmbati ,
Welcome back!
In Deep Learning you have weights (W) and biases (b) at each layer. These W and b are matrices and vectors respectively. The size of these objects depends on the number of neurons of each layer.
How are these parameters calculated?
At the beginning of the training, the W matrices are initialized with random values. There are several algorithms that can be used for this initialization. The bias vectors can be initialized with zeros.
Once the training start, these W and b parameters will be updated with new values by back propagation after each cycle.
So that’s it, that’s how parameters W and b are updated.
Please share any follow up question.
Thanks,
Juan.