sorry, we can't preview this file
BASS performance on synthetic data with pattern noise.
(a) The true probabilities of the motifs (yellow dots) and probabilities estimated (green dots) by our algorithm showing successful reconstruction of the dictionary for ϵp = 0.1, L = 40, 000. The crosses are low-probability motifs not identified by the algorithm. (b) The percentage of correctly placed partitions with increasing dataset size. Note that this measure differs from the measure used in Fig 2c (which additionally measures whether the correct symbol is decoded). This new measure is simpler to calculate when ϵp ≠ 0 since an extra alignment step is required to use the former measure when there are insertions and deletions. The corresponding optimal partitioning performance when the true dictionary is known is shown in dashed lines. In both panels, we use ϵb = 0.5, μ = 3, pd = 0.5.
(EPS)