Comparison with baseline methods

Do sophisticated algorithms always outperform simple baseline methods?


The above heatmap shows the ratio of the error of the primary algorithm to the error of a baseline algorithm (Identity or Uniform), across all datasets and various scales. The magnitude of the ratio is represented by the color in each cell.

  • Blue cells show the primary algorithm outperforming the baseline.
  • Red cells show the primary algorithm underperforming the baseline.
At small scale, there are some datasets, (e.g. BIDS-FJ) on which all algorithms are beaten by UNIFORM.