How can I define a constant vector layer in Caffe? - caffe

Question

I have a data with 10-d label vector, and I want to use a caffe model to make regression against these data with 10-d output. But now, I only want to check loss of some outputs (for example, 1, 3, 4, 5, 6-d of 10-d vector), so I define a layer with 5-d output at the bottom of the last output layer, But I'v no idea how to get corresponding 5-d label vector groundtruth, I think may be I can define a constant layer to indicate which entries I want get. Please help me if you have any ideas.
update: example
This is my original InnerProduct and Loss layer
layer {
name: "score"
type: "InnerProduct"
bottom: "fc7"
top: "score"
inner_product_param {
num_output: 10
weight_filler {
type: "gaussian"
std: 0.005
}
bias_filler {
type: "constant"
value: 1
}
}
}
layer {
name: "loss"
type: "EuclideanLoss"
bottom: "score"
bottom: "label"
top: "loss"
include {
phase: TRAIN
}
}
I care more about $n_1$ (like 1,3,4,5,6) entries of the 10-dimension output and their loss, so I want to fetch the loss of these entries, like
layer {
name: "score1"
type: "InnerProduct"
bottom: "fc7"
top: "score1"
inner_product_param {
num_output: 5 # n_1
weight_filler {
type: "gaussian"
std: 0.005
}
bias_filler {
type: "constant"
value: 1
}
}
}
layer {
name: "loss1"
type: "EuclideanLoss"
bottom: "score1"
bottom: "label"
top: "loss1"
include {
phase: TRAIN
}
}
How can I get score1 from score directly?

Answer 1

From what I interpret from your question, I think you would like to calculate the loss of the output layer w.r.t the labels for a regression model. But you would like to not bring some of the labels to the equation.
If my interpretation is true, being a regression model, I am expecting your new layer to be something similar to that of EuclidieanLayer. If it is so, the caffe_sub function in the layer could be replaced by the following code segment.
int arrayPos[5] = {1,3,4,5,6};
int count = 5;
Dtype *newBottom0=(Dtype*)malloc(sizeof(Dtype)*count);
Dtype *newBottom1=(Dtype*)malloc(sizeof(Dtype)*count);
for(int varI=0; varI<count; varI++)
{
newBottom0[varI] = (Dtype) bottom[0]->cpu_data()[arrayPos[varI]];
newBottom1[varI] = (Dtype) bottom[1]->cpu_data()[arrayPos[varI]];
}
caffe_sub( count, newBottom0, newBottom1, diff_.mutable_cpu_data());
free(newBottom0);
free(newBottom1);