caffe-innerproduct
usage:
layer {
name: "inn"
type: "InnerProduct"
bottom: "data"
top: "topinn"
param {
name: "w_hh"
}
inner_product_param {
num_output: 10
bias_term: false
weight_filler {
type: "uniform"
min: -0.8
max: 0.8
}
axis: 2 #default 1 # The first axis to be lumped into a single inner product computation;
# all preceding axes are retained in the output.
}
}
e.g.
data : 5 * 3 * 4 * 4
output: 10 axis: 2
=>
param: 10 * 16
datainn: 5 * 3 * 10