Skip to main content

Table 3 Term weighing schemes, taken from (Kralj et al. 2018), tested for decomposition of heterogeneous networks and their corresponding formulas

From: Py3plex toolkit for visualization and analysis of multilayer networks

Scheme Formula
tf f(t,d)
if-idf \(f(t, d) \cdot \log \left (\frac {|D|}{|\{d'\in D: t\in d'\}|}\right)\)
chi^2 \(f(t, d) \cdot \sum \limits _{c\in C}\frac {(P(t\wedge c)P(\neg t\wedge \neg c) - P(t\wedge \neg c) P(\neg t \wedge c))^{2}}{P(t)P(\neg t)P(c)P(\neg c)}\)
ig \(f(t, d)\cdot \sum \limits _{c\in C, c'\in \{c,\neg c\}t'\in \{t,\neg t\}}\left (P(t',c')\cdot \log \frac {P(t'\wedge c')}{P(t')P(c')}\right)\)
gr \(f(t, d)\cdot \sum \limits _{c\in C}\frac {{\sum \nolimits }_{c'\in \{c,\neg c\}}{\sum \nolimits }_{t'\in \{t,\neg t\}} \left (P(t',c')\cdot \log \frac {P(t'\wedge c')}{P(t')P(c')}\right)}{-{\sum \nolimits }_{c'\in \{c, \neg c\}}P(c)\cdot \log P(c)}\)
delta-idf \(f(t,d) \cdot \sum \limits _{c\in C}\left (\log \frac {|c|}{|\{d'\in D: d'\in c\wedge t\in d'\}|} - \log \frac {|\neg c|}{|\{d'\in D: d'\notin c\wedge t\notin d'\}|}\right)\)
rf \(f(t,d) \cdot \sum \limits _{c\in C}\log \left (2 + \frac {|\{d'\in D: d'\in c\wedge t\in d'\}|}{|\{d'\in D: d'\notin c\wedge t\notin d'\}|}\right)\)
bm25 \(f(t,d) \cdot \log \left (\frac {|D|}{|\{d'\in D: t\in d'\}|}\right)\cdot \frac {k + 1}{f(t, d) + k \cdot \left (1-b+b\cdot \frac {|d|}{\text {avgdl}}\right)}\)