

In [3]:

    
from spso2006 import Spso,eggHolder,rosen
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
%matplotlib inline
%load_ext autoreload
%autoreload 2

Eggholder function



In [4]:

    
nbRepet = 20
n = 20
nbEpoch=120
acceptableFitness = -959.6407



In [8]:

    
data_k_i = []
for n in [5,10,20,30]:
    for k in [2,3,4,5]:
        for i in range(nbRepet):
            pso = Spso(n,k=k,nbEvaluationMax=nbEpoch,acceptableFitness=acceptableFitness)
            pso.reset()
            pso.mainLoop()
            fit = pso.epochSummary['fitness']
            topo = pso.epochSummary['topology']
            indivs = pso.epochSummary['bestX']
            for epoch in range(len(fit)):
                x = indivs[epoch]
                data_k_i.append([n,k,i,epoch,fit[epoch],topo[epoch],x['x'],x['y'],x['z'],x['a'],x['b']])
#     k = 10000
#     for i in range(nbRepet):        
#         pso = Spso(n,k=k,nbEvaluationMax=nbEpoch,wl=0,wg=1,acceptableFitness=acceptableFitness)
#         pso.reset()
#         pso.mainLoop()
#         fit = pso.epochSummary['fitness']
#         indivs = pso.epochSummary['bestX']
#         for epoch in range(len(fit)):
#             x = indivs[epoch]
#             data_k_i.append([n,k,i,epoch,fit[epoch],topo[epoch],x['x'],x['y'],x['z'],x['a'],x['b']])









    



Process Worker-54904:
Traceback (most recent call last):
  File "/usr/lib/python3.4/multiprocessing/process.py", line 247, in _bootstrap
    util._run_after_forkers()






    



---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
<ipython-input-8-b964c2c46d01> in <module>()
      4         for i in range(nbRepet):
      5             pso = Spso(n,k=k,nbEvaluationMax=nbEpoch,acceptableFitness=acceptableFitness)
----> 6             pso.reset()
      7             pso.mainLoop()
      8             fit = pso.epochSummary['fitness']

/home/bchappet/Dropbox/THESE/workspace_python/SDNF/dnfpy/src/dnfpyUtils/optimisation/spso2006.py in reset(self)
     75 
     76         self.bestX = np.nan
---> 77         self.initFitness()
     78         self.fp = np.copy(self.f) #best fitness
     79         self.l = self.getBestNeighbor() #best previous best position in neighborhood amoung p

/home/bchappet/Dropbox/THESE/workspace_python/SDNF/dnfpy/src/dnfpyUtils/optimisation/spso2006.py in initFitness(self)
     86         self.f = np.ones((self.n)) * np.nan
     87         for i in range(self.n):
---> 88             self.manager.runEvaluation(i,self.indivToParams(self.x[i]))
     89             #f[i] = self.evaluate(self.indivToParams(self.p[i]))
     90         self.manager.finishWork()

/home/bchappet/Dropbox/THESE/workspace_python/SDNF/dnfpy/src/dnfpyUtils/optimisation/worker.py in runEvaluation(self, i, indiv)
     26         #request a worker for particle i. If no worker are available wait.
     27         #If a worker already works on i, return None, i will not be evaluated again
---> 28         worker = self.requestWorker(i)
     29         if worker:
     30             worker.evaluate(i,indiv)

/home/bchappet/Dropbox/THESE/workspace_python/SDNF/dnfpy/src/dnfpyUtils/optimisation/worker.py in requestWorker(self, i)
     55                     #we treat the queue instead of sleeping
     56                     else:
---> 57                         time.sleep(0.01)
     58         else:
     59             return

KeyboardInterrupt: 





    



  File "/usr/lib/python3.4/multiprocessing/util.py", line 127, in _run_after_forkers
    items = list(_afterfork_registry.items())
  File "/usr/lib/python3.4/weakref.py", line 191, in items
    yield k, v
KeyboardInterrupt



In [4]:

    
data =pd.DataFrame(data_k_i,columns=['n','k','i','epoch','f','topo','x','y','z','a','b'])



In [5]:

    
plt.figure(figsize=(16,10))
sns.tsplot(time='epoch',value='f',unit='i',condition='k',data=data[data.n==20])
plt.hlines(acceptableFitness,0,nbEpoch)









    Out[5]:





<matplotlib.collections.LineCollection at 0x7fa9b17ecc50>

Test multithread



In [60]:

    
nbRepet = 1
n = 20
nbEpoch=120
acceptableFitness = -959.6407
nbThread=8
function = eggHolder



In [62]:

    
data_k_i = []
for n in [20,]:
    print("n",n)
    k = 3
    for i in range(nbRepet):
        print("i",i)
        pso = Spso(n,k=k,nbEvaluationMax=nbEpoch,acceptableFitness=acceptableFitness,
                   nbThread=nbThread,evaluationFunc=eggHolder)
        pso.mainLoop()
        fit = pso.epochSummary['fitness']
        topo = pso.epochSummary['topology']
        indivs = pso.epochSummary['bestX']
        for epoch in range(len(fit)):
            x = indivs[epoch]
            data_k_i.append([n,k,i,epoch,fit[epoch],topo[epoch],x['x'],x['y'],x['z'],x['a'],x['b']])
#     k = 10000
#     for i in range(nbRepet):
#         pso = Spso(n,k=k,nbEvaluationMax=nbEpoch,wl=0,wg=1,acceptableFitness=acceptableFitness,
#                    nbThread=nbThread,evaluationFunc=eggHolder)
#         pso.mainLoop()
#         fit = pso.epochSummary['fitness']
#         indivs = pso.epochSummary['bestX']
#         for epoch in range(len(fit)):
#             x = indivs[epoch]
#             data_k_i.append([n,k,i,epoch,fit[epoch],topo[epoch],x['x'],x['y'],x['z'],x['a'],x['b']])









    



n 20
i 0
best Fitness  0 217.627802413
bestF : 217.62780241254634
bestX : {'b': -17.202300739662025, 'x': 398.73487869600922, 'a': 0.32790904944221211, 'z': -0.10390187421825914, 'y': -219.93274581933838}
best Fitness  2 -10.9344611516
bestF : -10.934461151556917
bestX : {'b': 12.499748602438821, 'x': -214.84204740407779, 'a': -12.549895256797615, 'z': 16.434265881454884, 'y': -445.0947285774073}
best Fitness  5 -211.654678663
bestF : -211.65467866346935
bestX : {'b': -12.321760343852457, 'x': 464.19279213676339, 'a': -16.770461457899888, 'z': 11.421579430050723, 'y': 404.77934942511536}
best Fitness  7 -233.780797867
bestF : -233.78079786749825
bestX : {'b': 8.908762384513814, 'x': 198.61580721284849, 'a': 9.6677190823688264, 'z': -0.33334894761712164, 'y': 214.15021959948217}
best Fitness  18 -367.789963407
bestF : -367.78996340680783
bestX : {'b': -0.5885312999212573, 'x': 179.67687297822454, 'a': -0.62221415543178082, 'z': 11.791295345577172, 'y': 298.725536851431}



In [63]:

    
data =pd.DataFrame(data_k_i,columns=['n','k','i','epoch','f','topo','x','y','z','a','b'])



In [66]:

    
plt.figure(figsize=(16,10))
#sns.tsplot(time='epoch',value='topo',unit='i',condition='k',data=data[data.n==20])
sns.tsplot(time='epoch',value='f',unit='i',condition='k',data=data[data.n==20])
#plt.hlines(acceptableFitness,0,nbEpoch)









    Out[66]:





<matplotlib.axes._subplots.AxesSubplot at 0x7feec4d888d0>



In [67]:

    
log = pso.log



In [68]:

    
f = np.array(log['f'])
p = np.array(log['p'])
x = np.array(log['x'])
v = np.array(log['v'])

fp = np.array(log['fp'])

l = np.array(log['l'])


lp = np.zeros_like(x)
for i in range(120):
    for j in range(20):
        lp[i,j,:] = p[i,l[i,j],:]



In [69]:

    
indiv = 10
plt.plot(fp[:,indiv])
plt.plot(f[:,indiv])
print(np.min(fp))









    



-959.640662614



In [72]:

    
indiv = 10
dim = 0
plt.plot(p[:,indiv,dim])
plt.plot(x[:,indiv,dim])
plt.plot(lp[:,indiv,dim])









    Out[72]:





[<matplotlib.lines.Line2D at 0x7feef0842278>]



In [ ]: