The average time per batch for 16 batches is 5,731/16 = 358.1875 hours.
Using the doubling rule, if the learning rate is R, then the time for the first batch is 500. The average time per batch for 2 batches is 500 x R. The averages time per batch for 4 batches is 500 x R^2 The average time per batch for 8 batches is 500 x R^3 The average time per batch for 16 batches is 500 x R^4
So 500 x R^4 = 358.1875 R^4 = 358.1875 / 500 =0.716375 R = fourth root of 0.716375 = 0.92 or 92%