Added gwr-inversion to requirements.txt.
Changed where gwr-inversion is imported.
Ensured that :func:np.ix_ receives arrays containing np.int32 objects.
Fixed a bug in the GPU version of the Verhulst model.
Attribute se of the output of :func:bd.estimate() is now a list instead of an array.
Simulation based standard errors and confidence intervals now use the estimated parameter as the initial condition for the optimisation procedure.
Improved how having a terminal state equal to 0 is handled by :func:bd.probability(method='gwasa').
Made it so an a value equal to 0 raises an exception in probability.gwasa.w_fun so that the high precision version of the function is invoked.
Updated how constraints are handled when differential evolution is used as an optimizer. Constraints are now always specified as dictionaries.
When polishing is used to enhance the output of differential evolution (which is the default action) it is done using method 'SLSQP' of :func:scipy.minimize() instead of method 'trust-constr' (which causes problems in other parts of the package).
Added the parameter 'stat' to the ABC algorithm so that 'mean' or 'median' can be used as a summary statistics for the posterior distribution.
Transition probability approximation methods 'gwa' and 'gwasa' now use a midpoint to anchor the linear approximation underlying their output.
Fixed a bug in :func:bd.forecast() related to displaying xticks.
Changed the default labels of x and y axis in bd.forecast().
Added the 'export' parameter to :func:bd.estimate() and :func:bd.forecast() which allows plots to be exported as tex files.
Removed a constant from the distance function in :func:bd.estimate(framework='abc').
Renamed :func:bdg.discrete_gpu to :func:bdg.discrete.
Renamed :func:bdg.probability_gpu to :func:bdg.probability.
Renamed attribute 'err' of :func:bd.estimate()'s output to 'val'.
Renamed argument 'num_bs_samples' of :func:bd.estimate() to 'num_samples'.
bpatch
commented
3 years ago