Continuing the discussion from About the Machine Learning Specialization category:

I am pretty sure I understand how to get the dimensions of parameters w and b here but what is explained here is completely different and confusing for me…

Is it explained wrong here?

can someone please explain this more clearly?

Hi @AnshumanKumar,

This is saying that for a specific layer, if s_{in} denotes the number of units in the layer, and s_{out} is the number of units in the next layer, then to go from this layer to the next layer, we apply a matrix W with dimensions s_{in}\times s_{out} and a vector b with dimension s_{out}. So for this particular layer, we are saying it has s_{in} inputs and s_{out} outputs.

For example in this network in the first layer, we would have s_{in} = 25 and s_{out} = 15, so to get from a 25-dimensional vector in the first layer to a 15-dimensional vector in the second layer, we apply a matrix W with dimension 25\times 15 and add a vector b of dimension 15.

I hope that helps, but feel free to ask for any additional clarification.

Best,

Alex

2 Likes

Thanks sir, the example cleared all the confusion

1 Like