DynamicNode¶
This class implements a node of the map. Each node uses a Hidden Markov Model to represent it’s dynamics. It can have any number of states and observed symbols.
This class only needs to know the number of states and symbols and the initial guess of the transition and emission probability matrices (A and B). It contains methods for updating the transition probability matrices and to compute the current probability of each state.
All the math is actually done in a lower level module _online_em, which is
written in C to improve speed. The HMM training and the state probability
estimations use the On Line algorithm described by Mongillo and Deneve.
After building the map, the method DynamicMap.map_iter() can be used to
get a list of coordinates with the correspondent nodes.
API¶
Next, find the public API of the class, the full list of methods (including the _-prefixed ones) can be found in DynamicNode - All Members.