In discrete memoryless channels what is meant by output probabilities?

In my book it is mentioned that
P[Y]= P(X)* P(Y/X)
Where P[Y] is the output probability matrix, P(X)=input probability matrix, P(Y/X)= channel matrix.
I understand the physical significance of input probabilities and channel matrix probabilities because any input X can be generated which inturn determines the probabilities of generation of Y. But what do we mean by output probabilities? My question: is there any physical significance of output probabilities without defining which input is producing them?