Skip to content

Commit

Permalink
Create scrank.py
Browse files Browse the repository at this point in the history
  • Loading branch information
@Raschild
Raschild committed Dec 12, 2015
1 parent 901a6f2 commit bde6a4b
Showing 1 changed file with 301 additions and 0 deletions.
301 changes: 301 additions & 0 deletions python/scrank.py
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,301 @@
from __future__ import division
from math import log10
import numpy as np
import json
import sys

how_to = "Please give: \n" \
"--run noweight --dumpfactor <float number> --numtasks <total tasks> --lastcompleted <position of last completed task>" \
" \nOR\n" \
"--run weighted1 || weighted2 --dumpfactor <float number> --numtasks <total tasks> --completed <taskNumber1> = <value1>, <taskNumber2> = <value2>" \
" \nOR\n" \
"--run weighted3 --dumpfactor <float number> --numtasks <total tasks> --completed <taskNumber1> = <value1>, <taskNumber2> = <value2> --time <taskNumber1> = <value1>, <taskNumber2> = <value2>"


## Consider:
# 1. completion rate of task itself and even distribution of out-edges-weight among rest of tasks
# 1. completion rates of other tasks
# 2. time expiration rates of other tasks
def WeightedSCRank3(dumping_factor, num_tasks, completion_rates, time_rates):

sum = 0
## Find the total remaining time for all the tasks
for time in time_rates:
sum += time_rates[time]

## For each task compute the **IMPORTANCE** of time
for time in time_rates:
time_rates[time] = abs(log10(time_rates[time]/sum))


Vprev = np.array([1/num_tasks]*num_tasks)
Tax = np.array([(1-dumping_factor)/num_tasks]*num_tasks)

rows = []
for task in range(0,num_tasks):
## node 0 is the value 0+1 inside the quotas dictionary
## every outgoing edge should have the value of the completed tasks multiplied by
same_node_weight_final = 1-completion_rates[task+1]
other_node_weight_final = completion_rates[task+1]/(num_tasks-1)

sum = 0
row = []
##
for curr_pos in range(0,num_tasks):
if curr_pos != task:
other_node_weight = other_node_weight_final * (1-completion_rates[curr_pos+1]) * time_rates[curr_pos+1]
row.append(other_node_weight)
sum += other_node_weight
else:
same_node_weight = same_node_weight_final * (1-completion_rates[curr_pos+1]) * time_rates[curr_pos+1]
row.append(same_node_weight)
sum += same_node_weight
row[:] = [x/sum for x in row]
rows.append(row)


Transition_matrix = np.array(rows)
Transition_matrixT = np.transpose(Transition_matrix)

error=float('inf')
num_iters = 0
threshold = 0.1
Vnext = None
while (error > threshold):
Vnext = np.dot(dumping_factor*Transition_matrixT,Vprev) + Tax
error = np.linalg.norm((Vnext-Vprev))
Vprev = Vnext
num_iters += 1


print("Taxation: ", Tax)
print("Transition Matrix Transpose: ", Transition_matrixT)
print("Vector Computed: ", Vnext)
print(np.sum(Vnext))
print("Error: ", error)
print("Iterations Performed: ", num_iters)

return Vnext



## Consider:
# 1. completion rate of task itself and even distribution of out-edges-weight among rest of tasks
# 2. completion rates of other tasks
def WeightedSCRank2(dumping_factor, num_tasks, completion_rates):

# dumping_factor = 1
# num_tasks = 3
# completion_rates = {1:0.1,2:0.5,3:0.8}

Vprev = np.array([1/num_tasks]*num_tasks)
Tax = np.array([(1-dumping_factor)/num_tasks]*num_tasks)

rows = []
for task in range(0,num_tasks):
## node 0 is the value 0+1 inside the quotas dictionary
## every outgoing edge should have the value of the completed tasks multiplied by
same_node_weight_final = 1-completion_rates[task+1]
other_node_weight_final = completion_rates[task+1]/(num_tasks-1)

sum = 0
row = []
for curr_pos in range(0,num_tasks):
if curr_pos != task:
other_node_weight = other_node_weight_final * (1-completion_rates[curr_pos+1])
row.append(other_node_weight)
sum += other_node_weight
else:
same_node_weight = same_node_weight_final * (1-completion_rates[curr_pos+1])
row.append(same_node_weight)
sum += same_node_weight
row[:] = [x/sum for x in row]
rows.append(row)


Transition_matrix = np.array(rows)
Transition_matrixT = np.transpose(Transition_matrix)

error=float('inf')
num_iters = 0
threshold = 0.1
Vnext = ""
while (error > threshold):
Vnext = np.dot(dumping_factor*Transition_matrixT,Vprev) + Tax
error = np.linalg.norm((Vnext-Vprev))
Vprev = Vnext
num_iters += 1

print("Taxation: ", Tax)
print("Transition Matrix Transpose: ", Transition_matrixT)
print("Vector Computed: ", Vnext)
print(np.sum(Vnext))
print("Error: ", error)
print("Iterations Performed: ", num_iters)

return Vnext

## Consider:
# 1. completion rate of task itself and even distribution of out-edges-weight among rest of tasks
def WeightedSCRank1(dumping_factor, num_tasks, completion_rates):

# dumping_factor = 1
# num_tasks = 10
# completion_rates = {1:0.05,2:0.5,3:0.2,4:0.5,5:0.9,6:0.7,7:0.8,8:0.9,9:0.1,10:0.95}


Vprev = np.array([1/num_tasks]*num_tasks)
Tax = np.array([(1-dumping_factor)/num_tasks]*num_tasks)
rows = []
for task in range(0,num_tasks):
## node 0 is the value 0+1 inside the quotas dictionary
## every outgoing edge should have the value of the completed tasks multiplied by
same_node_weight = 1-completion_rates[task+1]
other_node_weight = completion_rates[task+1]/(num_tasks-1)
row = [other_node_weight if curr_pos != task else same_node_weight for curr_pos in range(0,num_tasks)]
rows.append(row)

Transition_matrix = np.array(rows)
Transition_matrixT = np.transpose(Transition_matrix)

error=float('inf')
num_iters = 0
threshold = 0.1
Vnext = ""
while (error > threshold):
Vnext = np.dot(dumping_factor*Transition_matrixT,Vprev) + Tax
error = np.linalg.norm((Vnext-Vprev))
Vprev = Vnext
num_iters += 1

print("Taxation: ", Tax)
print("Transition Matrix Transpose: ", Transition_matrixT)
print("Vector Computed: ", Vnext)
print(np.sum(Vnext))
print("Error: ", error)
print("Iterations Performed: ", num_iters)

return Vnext


# Consider:
# 1. Last completed task -- No incoming edge for this task from the rest of the nodes
def NoWeightSCRank(dumping_factor, num_tasks, last_completed_task):
# dumping_factor = 0.85
Vprev = np.array([1/num_tasks]*num_tasks)
Tax = np.array([(1-dumping_factor)/num_tasks]*num_tasks)
rows = []
position = last_completed_task-1
for task in range(0,num_tasks):
if task != position:
## include all the tasks; exclude the last completed
row = [1/(num_tasks-1) if curr_pos!=position else 0 for curr_pos in range(0,num_tasks)]
else:
row = [1/num_tasks for curr_pos in range(0,num_tasks)]
rows.append(row)

Transition_matrix = np.array(rows)
Transition_matrixT = np.transpose(Transition_matrix)

error=float('inf')
num_iters = 0
threshold = 0.1
Vnext = ""
while (error > threshold):
Vnext = np.dot(dumping_factor*Transition_matrixT,Vprev) + Tax
error = np.linalg.norm((Vnext-Vprev))
Vprev = Vnext
num_iters += 1


print("Taxation: ", Tax)
print("Transition Matrix Transpose: ", Transition_matrixT)
print("Vector Computed: ", Vnext)
print(np.sum(Vnext))
print("Error: ", error)
print("Iterations Performed: ", num_iters)

return Vnext



def CallNoWeightSCRank(sys_args):
if len(sys_args)>4 and sys_args[3] == "--dumpfactor":
dumping_factor = float(sys_args[4])
if len(sys_args)>6 and sys_args[5] == "--numtasks":
num_tasks = int(sys_args[6])
if len(sys_args)>7 and sys_args[7] == "--lastcompleted":
last_completed = int(sys_args[8])
NoWeightSCRank(dumping_factor, num_tasks, last_completed)
else:
print("Plase give the number of the last completed task")
else:
print("Please give the number of tasks")
else:
print("Please give the dumping factor")


def CallWeightedSCRank12(sys_args, alg_name):
if len(sys_args)>4 and sys_args[3] == "--dumpfactor":
dumping_factor = float(sys_args[4])
if len(sys_args)>6 and sys_args[5] == "--numtasks":
num_tasks = int(sys_args[6])
if len(sys_args)>7 and sys_args[7] == "--completed":
i=8
completion_rates = {}
while i<len(sys_args):
task = (sys_args[i].split("="))
completion_rates[int(task[0])] = float(task[1])
i+=1
if alg_name == "weighted1":
print WeightedSCRank1(dumping_factor, num_tasks, completion_rates)
elif alg_name == "weighted2":
print WeightedSCRank2(dumping_factor, num_tasks, completion_rates)
else:
print("Please give the completed rates of the tasks")
else:
print("Please give the total number of the tasks")


def CallWeightedSCRank3(sys_args):
if len(sys_args)>4 and sys_args[3] == "--dumpfactor":
dumping_factor = float(sys_args[4])
if len(sys_args)>6 and sys_args[5] == "--numtasks":
num_tasks = int(sys_args[6])
if len(sys_args)>7 and sys_args[7] == "--completed":
i=8
completion_rates = {}
while i<len(sys_args) and sys_args[i] != "--time":
task = (sys_args[i].split("="))
completion_rates[int(task[0])] = float(task[1])
i+=1
if i==len(sys_args) or len(sys_args)<i+2:
print("Please give the remaining time of the tasks in Minutes OR in the same time unit metric(e.g. seconds, hour)")
else:
i+=1
time_rates = {}
while i<len(sys_args):
task = (sys_args[i].split("="))
time_rates[int(task[0])] = float(task[1])
i += 1
print(WeightedSCRank3(dumping_factor, num_tasks, completion_rates, time_rates))
else:
print("Please give the completed rates of the tasks")
else:
print("Please give the total number of the tasks")


if __name__ == "__main__":

if len(sys.argv)>1 and sys.argv[1] == "--run":
run_mode = sys.argv[2]
if run_mode == "noweight":
CallNoWeightSCRank(sys.argv)
elif run_mode == "weighted1":
CallWeightedSCRank12(sys.argv, "weighted1")
elif run_mode == "weighted2":
CallWeightedSCRank12(sys.argv, "weighted2")
elif run_mode == "weighted3":
CallWeightedSCRank3(sys.argv)
else:
print(how_to)

0 comments on commit bde6a4b

Please sign in to comment.