local_progress_reached
python
local_progress_reached(*, start_route_value: float, best_successor_route_value: float, reached_route_value: float, gap_fraction: float) -> boolDecide whether one episode made local progress along the route.
Progress = reaching a cell whose route value clears a bar set gap_fraction of the way from the start cell to its best stored successor. When no better successor is known (best_successor_route_value <= start_route_value) the bar collapses to the start value, so merely not regressing counts as progress — such a cell has no better next hop to learn and should read as already mastered rather than as a frontier.
Parameters:
start_route_value(float) – Route-utility cell value at the episode's start cell.best_successor_route_value(float) – Highest route-utility cell value among the start cell's known successors.reached_route_value(float) – Highest route-utility cell value the episode reached.gap_fraction(float) – Fraction of the start-to-best-successor gap that must be covered (α).
Returns:
bool–Truewhen the episode cleared the bar.