Hello all,

Say I have an experiment to find the relationship of speed as a function of time and distance. We all know that speed = distance / time.

Now, if I were to take say 10,000 realistic observations (i.e., these have a non-zero probability of error) and train a regressor to estimate speed. Does it offer any value to do so?

My line of thought is this, can it be that: distance = speed * time + O(1)? By extrapolation of thought, say we had more complex relationships such as the use of SINR, bandwidth, FEC, MIMO rank, etc to compute the sum-rate capacity. Would using ML find a surrogate relationship that is slightly different from Shannon's formula be worthwhile?

Similar questions and discussions