Surely the difference is below the noise level of whatever data were used to get these figures? I make it about 0.16 millibits of Kullback-Leibler divergence, and upwards of 10,000 data points to have any chance of seeing the difference experimentally.
Surely the difference is below the noise level of whatever data were used to get these figures? I make it about 0.16 millibits of Kullback-Leibler divergence, and upwards of 10,000 data points to have any chance of seeing the difference experimentally.