THIS IS A TEST INSTANCE. ALL YOUR CHANGES WILL BE LOST!!!!
...
number of passes over the data = epochs * number_of_iterations
Increasing epochs
while reducing number_of_iterations
to maintain the same number of passes can result in significantly faster training because there is less hopping. However, but there may be an impact on convergence because you are visiting the same examples more than once in an per iteration , which violates logical sequential SGD, which theoretically has the best convergence efficiency.
...