pareyesv / landscape_steiner Goto Github PK

Follow cookiecutter datascience conventions

Also format the code

sdasdasd

See also:
https://steinlib.readthedocs.io

Cell 14 in steiner_timing

change to

X, Y = [x for x,y in varyNterm], [y[-1] for x,y in varyNterm]

Ref:

• modify create_graph en steiner_utils.py
  

  landscape_steiner/steiner_utils.py

  Lines 26 to 40 in 824425e

  	def create_graph(box_size, terminals):
  	''' Initialize 2D grid graph with terminals added as node attributes '''
  	# create square grid graph
  	grid_1d = range(1, int(box_size))
  	G = nx.grid_2d_graph(grid_1d, grid_1d)
  	# set terminal as attribute
  	attribute_dictonary = {}
  	for point in terminals:
  	attribute_dictonary.update({point: {'terminal': 1}})
  	nx.set_node_attributes(G, attribute_dictonary)
  	return G

• or create a new function to add support to non-grid graphs

Add function to compute box_size to be passed as argument here:

box_size must be implemented for non-grid graphs

There is a much nicer Landscape Function which doesn't require solving PDEs, essentially

fLandscape = 1/(1+V)

where V = V(x,y) is the potential given by the sum of gaussians.

This makes the code faster, easier to understand, deals with the boundaries of the domain a lot better, and makes the range of the cutoff 'alpha' finite.

Example:
total_budget_limit-01 -> total_budget_limit-01-01

Thus it will be able to assign budget limit to multiple (consecutive) periods (as in Faria2020).
For example total_budget_limit-01-03 will correspond to the total budget limit for periods 1,2, and 3.