GBA uses forget rate (α) to compute the Repeat (R) and Non-Repeat (NR) matrices. These two matrices are used to filter the vertices
constructed for similar letter pairs when they are in the same window by chance (i.e. false positives).
R and NR are 20X20 matrices that show statistical information
on frequencies of pairs of letters in repeat and non-repeat regions respectively. The values R(γi,γj)
and R(γi,γj) represent average probabilities that γi and γj
appear together in a repeat region and non-repeat region, respectively. Forget rate (α) is used as a parameter to calculate R
and NR ( 0 ≤ α ≤ 1). In general, it is used in various domains to capture the correlation between two objects
that are members of an ordered list of objects based on their spacial distance in that list. A default value of α is 0.95.
t1 is the length of the sliding window w. GBA uses w to construct the graph of vertices pair. The default
value of t1 is 15.
t2 and t3 are two distance threshold measure in GBA. Once, the graph of vericies pair is constructed,
GBA incerts edges btween vertices. An edge is added between two vertices (i, j) and (k, m) if s(i)s(k) and s(j)s(m)
do not conflict with each other. This property is enforced by introducing distance threshold t2 and t3.
t2 specifies the maximum number of insertions and deletions between similar repeats. t3 specifies the
maximum distance between letters in cryptic repeats. The default values of t2 and t3 are 3 and 5 respectively.