A lexical approach for identifying behavioural action sequences
Fig 1
The generative model from motifs to behavioural output.
(a) Motif templates are fixed sequences of elementary locomotor episodes (labeled a, b and c in this example). The observed behavioural output is generated from motif templates drawn sequentially from a dictionary. An instantiation of a template may “mutate” by insertions (red) or deletions (blue), which then generates the observed output as shown in panel (b). (b) The generative process from a motif template c1 c2…cl to instantiation to observed output
. (c) The unsupervised inference procedure (BASS) first learns a dictionary of motifs and then segments (vertical bars) the observed behavioural output y1, y2, … into the most likely sequence of motifs m1, m2, … from the dictionary that generated it.