I have 30 decaying traces. Each of them has slight different rate of change. I want to model (classical or RNN) based on 5 initial rate of changes after every 1 minute. So I have R1,R2,R3,R4,R5 till first 5 minutes, and then want to predict the rest of the decaying trace. For training we can use whole traces, at inference time the input will be 5 rates. What are the possible ways to model this?