We start by loading the stored files. For this, we load the evaluation library
In [1]:
# General libraries
import os
import pickle
import numpy as np
# Plot, in nb, only when .show() is called
import matplotlib.pyplot as plt
%matplotlib notebook
plt.ioff()
# Personal libraries
import tools.evaluation as ev
import tools.plot as pt
Before evaluating the simulaion results, we scan the folders and create the variables where we will store the overlaps. Note that each type of simulations has a six letters heading. For example, for Kinetic Capacity simulation the heading is KinCap
In [2]:
path = '/home/sartori/Data/memexp'
# path = '/Users/pablo/Data/memexp'
dir_list = [ path + '/' + x for x in os.listdir(path) if x[0:9]=='KinSeqLen']
dir_list.sort()
We can now read the files and store the overlaps over time. We store them in a dictionary which assigns to the tuple (k,p) a list with all the overlaps over time (this implicitly assumes that all other parameters, such as network size or temperature, were kept fixed equal)
In [26]:
data = {}
for d in dir_list:
# Load data
net = pickle.load( open(d + '/net.pkl','rb') )
overlaps = np.load(d + '/overlaps.npy')
# Create the key to store data
key = (net.k, net.p)
data.setdefault(key,[]).append(overlaps)
print d
# Generate plot
#pt.plot_overlaps(overlaps, show = False, name = os.path.basename(d) )
We can look at the values of (p,k) that we have loaded
In [24]:
data.keys()
Out[24]:
We can plot the overlaps for one example
In [22]:
key = (5,15)
test_overlaps = data[key][0]
pt.plot_overlaps(test_overlaps[:,:])
Out[22]:
We can also check the score of the retrieved pattern
In [23]:
ev.my_score(key, data[key], cutoff = 0.5, time_retrieved = 10)
Out[23]:
Analysis
In [44]:
reload(ev)
xyz = []
for key in data.iterkeys():
wowo = ev.my_score(key, data[key], cutoff = 0.75, time_retrieved = 15)
xyz.append( (key[0], key[1], wowo) )
XYZ = np.asarray(xyz)
XY = XYZ[:,1:]
Out[44]:
In [12]:
N_600_k_6 = np.array([[ 6. , 0.95369048],
[ 7. , 0.94718831],
[ 8. , 0.95378571],
[ 9. , 0.94807143],
[ 10. , 0.93169264],
[ 11. , 0.95302381],
[ 12. , 0.95964286],
[ 13. , 0.95659524],
[ 14. , 0.9325873 ],
[ 15. , 0.93683117],
[ 16. , 0.92232305],
[ 17. , 0.92747619],
[ 18. , 0.93204978],
[ 19. , 0.92041486],
[ 20. , 0.91522746],
[ 21. , 0.90003422],
[ 22. , 0.92825216],
[ 23. , 0.95593074],
[ 24. , 0.95315476],
[ 25. , 0.87068209],
[ 26. , 0.92831513],
[ 27. , 0.91080292],
[ 28. , 0.88504762],
[ 29. , 0.90925216],
[ 30. , 0.87813409],
[ 31. , 0.87394481],
[ 32. , 0.85943898],
[ 33. , 0.87564141],
[ 34. , 0.8303058 ],
[ 35. , 0.82828851],
[ 36. , 0.79966836],
[ 37. , 0.82360065],
[ 38. , 0.85686888],
[ 39. , 0.84079942],
[ 40. , 0.8664228 ],
[ 41. , 0.84274459],
[ 42. , 0.7854203 ],
[ 43. , 0.7805369 ],
[ 44. , 0.80515279],
[ 45. , 0.79861724]])
N_300_k_5 = np.array([[ 5. , 0.95088889],
[ 6. , 0.97050794],
[ 7. , 0.91151587],
[ 8. , 0.9357619 ],
[ 9. , 0.96209127],
[ 10. , 0.9257619 ],
[ 11. , 0.9114127 ],
[ 12. , 0.9205119 ],
[ 13. , 0.91901462],
[ 14. , 0.88436383],
[ 15. , 0.91598016],
[ 16. , 0.90003719],
[ 17. , 0.86954906],
[ 18. , 0.87684491],
[ 19. , 0.83830397],
[ 20. , 0.83805614],
[ 21. , 0.84352797],
[ 22. , 0.84136762],
[ 23. , 0.81173737],
[ 24. , 0.82705898],
[ 25. , 0.79449098],
[ 26. , 0.73145088],
[ 27. , 0.7906142 ],
[ 28. , 0.74260462],
[ 29. , 0.7401806 ],
[ 30. , 0.71375386],
[ 31. , 0.69474206],
[ 32. , 0.66053139],
[ 33. , 0.67109002],
[ 34. , 0.62481405]])
N_300_k_7 = np.array([[ 7. , 0.82206571],
[ 8. , 0.81966461],
[ 9. , 0.80958555],
[ 10. , 0.83364291],
[ 11. , 0.82373521],
[ 12. , 0.82612116],
[ 13. , 0.79994331],
[ 14. , 0.79163858],
[ 15. , 0.79444012],
[ 16. , 0.79028216],
[ 17. , 0.81237318],
[ 18. , 0.7482349 ],
[ 19. , 0.7502307 ],
[ 20. , 0.75578011],
[ 21. , 0.73713936],
[ 22. , 0.73331323],
[ 23. , 0.76652548],
[ 24. , 0.72488869],
[ 25. , 0.67307359],
[ 26. , 0.6778688 ],
[ 27. , 0.68594744],
[ 28. , 0.68337721],
[ 29. , 0.59460317],
[ 30. , 0.58188119],
[ 31. , 0.54068741],
[ 32. , 0.58741868],
[ 33. , 0.5091241 ],
[ 34. , 0.60376277],
[ 35. , 0.45519889],
[ 36. , 0.43525042]])
We now generate a plot
In [14]:
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax1.scatter(N_300_k_5[:,0], N_300_k_5[:,1], s=35, c='g', marker="o", label='$N=300,\;k=5$')
ax1.scatter(N_300_k_7[:,0], N_300_k_7[:,1], s=35, c='b', marker="o", label='$N=300,\;k=7$')
ax1.scatter(N_600_k_6[:,0], N_600_k_6[:,1], s=35, c='r', marker="s", label='$N=600,\;k=6$')
plt.xlabel('stored patterns, $p$',fontdict={'fontsize':20})
plt.ylabel('sequence retrieval score',fontdict={'fontsize':20})
plt.axis([0, 50, 0, 1])
plt.legend(loc='upper right');
plt.show()
We start by loading the stored files. For this, we load the evaluation library
In [1]:
# General libraries
import os
import pickle
import numpy as np
# Plot, in nb, only when .show() is called
import matplotlib.pyplot as plt
%matplotlib notebook
plt.ioff()
# Personal libraries
import tools.evaluation as ev
import tools.plot as pt
Before evaluating the simulaion results, we scan the folders and create the variables where we will store the overlaps. Note that each type of simulations has a six letters heading. For example, for Kinetic Capacity simulation the heading is EnCapCorr
In [ ]:
path = '/home/sartori/Data/memexp'
# path = '/Users/pablo/Data/memexp'
dir_list = [ path + '/' + x for x in os.listdir(path) if x[0:9]=='EnCapCorr']
dir_list.sort()
We can now read the files and store the overlaps over time. We store them in a dictionary which assigns to the tuple (k,p) a list with all the overlaps over time (this implicitly assumes that all other parameters, such as network size or temperature, were kept fixed equal)
In [26]:
data = {}
for d in dir_list:
# Load data
overlaps = np.load(d + '/overlaps.npy')
# Create the key to store data
p = int(d.split("_")[1][1:])
c = int(d.split("_")[3][1:])
key = (c, p)
# Store the data
data.setdefault(key,[]).append(overlaps)
print d
# Generate plot
#pt.plot_overlaps(overlaps, show = False, name = os.path.basename(d) )
Analysis for the case of correlated patterns
In [ ]:
xyz = []
for key in data.iterkeys():
k = 15
p = key[1]
wowo = ev.my_score((k, p), data[key], cutoff = 0.75, time_retrieved = 15)
xyz.append( (k, c, wowo) )
XYZ = np.asarray(xyz)
XY = XYZ[:,1:]
Collect the data form the simulations tun in the cluster
In [4]:
Ec30k7 = np.array([[ 7.00000000e+00, 8.45969336e-01],
[ 8.00000000e+00, 8.17170274e-01],
[ 9.00000000e+00, 7.35082126e-01],
[ 1.00000000e+01, 6.38407703e-01],
[ 1.10000000e+01, 5.99932900e-01],
[ 1.20000000e+01, 4.64281205e-01],
[ 1.30000000e+01, 4.01477273e-01],
[ 1.40000000e+01, 2.31383690e-01],
[ 1.50000000e+01, 1.87150794e-01],
[ 1.60000000e+01, 6.57251082e-02],
[ 1.70000000e+01, 4.19706960e-02],
[ 1.80000000e+01, 6.39527417e-02],
[ 1.90000000e+01, 2.32622655e-02],
[ 2.00000000e+01, 3.22499722e-02],
[ 2.10000000e+01, 2.22720842e-02],
[ 2.20000000e+01, 1.72702020e-02],
[ 2.30000000e+01, 1.31997863e-02],
[ 2.40000000e+01, 1.10049020e-02],
[ 2.50000000e+01, 1.88658009e-02],
[ 2.60000000e+01, 2.11847874e-02],
[ 2.70000000e+01, 1.30555556e-02],
[ 2.80000000e+01, 1.26439394e-02],
[ 2.90000000e+01, 9.88442113e-03]]) # cutoff = 0.75, dt = 15
Ec20k7 = np.array([[ 7. , 0.725 ],
[ 8. , 0.77942063],
[ 9. , 0.71456349],
[ 10. , 0.74063492],
[ 11. , 0.73959921],
[ 12. , 0.67109127],
[ 13. , 0.71756349],
[ 14. , 0.69521429],
[ 15. , 0.64073413],
[ 16. , 0.63285714],
[ 17. , 0.60983766],
[ 18. , 0.54068987],
[ 19. , 0.49763492],
[ 20. , 0.4125 ],
[ 21. , 0.37652778],
[ 22. , 0.38134921],
[ 23. , 0.26017857],
[ 24. , 0.22642857],
[ 25. , 0.15038961],
[ 26. , 0.10321429],
[ 27. , 0.1325 ],
[ 28. , 0.05535714],
[ 29. , 0.03160714]])
Ec15k7 = np.array([[ 7. , 0.60678571],
[ 8. , 0.59357143],
[ 9. , 0.59956349],
[ 10. , 0.62821429],
[ 11. , 0.56107143],
[ 12. , 0.64053571],
[ 13. , 0.58785714],
[ 14. , 0.53742063],
[ 15. , 0.58339286],
[ 16. , 0.57498016],
[ 17. , 0.58402778],
[ 18. , 0.55321429],
[ 19. , 0.607 ],
[ 20. , 0.58910714],
[ 21. , 0.56714286],
[ 22. , 0.49535714],
[ 23. , 0.53660714],
[ 24. , 0.47593254],
[ 25. , 0.41259921],
[ 26. , 0.44303571],
[ 27. , 0.35325397],
[ 28. , 0.41321429],
[ 29. , 0.28053571]])
Ec10k7 = np.array([[ 7. , 0.3875 ],
[ 8. , 0.45321429],
[ 9. , 0.48089286],
[ 10. , 0.48535714],
[ 11. , 0.46732143],
[ 12. , 0.4575 ],
[ 13. , 0.46589286],
[ 14. , 0.51777778],
[ 15. , 0.49517857],
[ 16. , 0.47160714],
[ 17. , 0.40428571],
[ 18. , 0.4975 ],
[ 19. , 0.44160714],
[ 20. , 0.46607143],
[ 21. , 0.555 ],
[ 22. , 0.48785714],
[ 23. , 0.48160714],
[ 24. , 0.43142857],
[ 25. , 0.47910714],
[ 26. , 0.51089286],
[ 27. , 0.48589286],
[ 28. , 0.45714286],
[ 29. , 0.41 ]])
Ec5k7 = np.array([[ 7. , 0.36142857],
[ 8. , 0.38 ],
[ 9. , 0.35285714],
[ 10. , 0.38 ],
[ 11. , 0.34875 ],
[ 12. , 0.4 ],
[ 13. , 0.34321429],
[ 14. , 0.34920635],
[ 15. , 0.35732143],
[ 16. , 0.36857143],
[ 17. , 0.33428571],
[ 18. , 0.37 ],
[ 19. , 0.35428571],
[ 20. , 0.39206349],
[ 21. , 0.36464286],
[ 22. , 0.35428571],
[ 23. , 0.37142857],
[ 24. , 0.35142857],
[ 25. , 0.40303571],
[ 26. , 0.39732143],
[ 27. , 0.33589286],
[ 28. , 0.35321429],
[ 29. , 0.31857143]])
Ec0k7 = np.array([[ 7. , 0.24571429],
[ 8. , 0.28571429],
[ 9. , 0.30607143],
[ 10. , 0.27285714],
[ 11. , 0.29571429],
[ 12. , 0.29428571],
[ 13. , 0.26428571],
[ 14. , 0.28285714],
[ 15. , 0.27714286],
[ 16. , 0.28285714],
[ 17. , 0.28142857],
[ 18. , 0.28857143],
[ 19. , 0.26714286],
[ 20. , 0.28285714],
[ 21. , 0.27285714],
[ 22. , 0.28142857],
[ 23. , 0.26035714],
[ 24. , 0.28142857],
[ 25. , 0.29571429],
[ 26. , 0.25 ],
[ 27. , 0.28285714],
[ 28. , 0.25142857],
[ 29. , 0.28571429]])
Ec0k14 = np.array([[ 14. , 0.14857143],
[ 15. , 0.13571429],
[ 16. , 0.15 ],
[ 17. , 0.12785714],
[ 18. , 0.14785714],
[ 19. , 0.13785714],
[ 20. , 0.13071429],
[ 21. , 0.14285714],
[ 22. , 0.15357143],
[ 23. , 0.135 ],
[ 24. , 0.13642857],
[ 25. , 0.16642857],
[ 26. , 0.13714286],
[ 27. , 0.13714286],
[ 28. , 0.14571429],
[ 29. , 0.12571429],
[ 30. , 0.13785714],
[ 31. , 0.145 ],
[ 32. , 0.135 ],
[ 33. , 0.135 ],
[ 34. , 0.14071429],
[ 35. , 0.135 ],
[ 36. , 0.135 ]])
Kc30k7 = np.array([[ 7.00000000e+00, 3.21269841e-01],
[ 8.00000000e+00, 2.62678571e-01],
[ 9.00000000e+00, 2.38571429e-01],
[ 1.00000000e+01, 1.68571429e-01],
[ 1.10000000e+01, 1.26300366e-01],
[ 1.20000000e+01, 9.28571429e-02],
[ 1.30000000e+01, 7.71428571e-02],
[ 1.40000000e+01, 5.00000000e-02],
[ 1.50000000e+01, 2.85714286e-02],
[ 1.60000000e+01, 1.10714286e-02],
[ 1.70000000e+01, 7.38095238e-03],
[ 1.80000000e+01, 2.19780220e-03],
[ 1.90000000e+01, 0.00000000e+00],
[ 2.00000000e+01, 2.33766234e-03],
[ 2.10000000e+01, 0.00000000e+00],
[ 2.20000000e+01, 0.00000000e+00],
[ 2.30000000e+01, 1.25000000e-03],
[ 2.40000000e+01, 0.00000000e+00],
[ 2.50000000e+01, 0.00000000e+00],
[ 2.60000000e+01, 0.00000000e+00],
[ 2.70000000e+01, 0.00000000e+00],
[ 2.80000000e+01, 0.00000000e+00],
[ 2.90000000e+01, 0.00000000e+00]]) # cutoff = 0.75, dt = 15
In [5]:
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax1.scatter(Ec30k7[:,0], Ec30k7[:,1], s=35, c='k', marker="o", label='Energetic, $c=30,\;k=7$')
ax1.scatter(Ec20k7[:,0], Ec20k7[:,1], s=35, c='0.6', marker="o", label='Energetic, $c=20,\;k=7$')
ax1.scatter(Ec15k7[:,0], Ec15k7[:,1], s=35, c='r', marker="o", label='Energetic, $c=15,\;k=7$')
ax1.scatter(Ec10k7[:,0], Ec10k7[:,1], s=35, c='g', marker="o", label='Energetic, $c=10,\;k=7$')
ax1.scatter(Ec5k7[:,0], Ec5k7[:,1], s=35, c='y', marker="o", label='Energetic, $c=5,\;k=7$')
ax1.scatter(Ec0k7[:,0], Ec0k7[:,1], s=35, c='b', marker="o", label='Energetic, $c=0,\;k=7$')
ax1.scatter(Ec0k14[:,0], Ec0k14[:,1], s=35, c='w',marker="o", label='Energetic, $c=0,\;k=14$')
ax1.scatter(Kc30k7[:,0], Kc30k7[:,1], s=35, c='b', marker="s", label='Kinetic, $c=30,\;k=7$')
plt.xlabel('stored patterns, $p$',fontdict={'fontsize':20})
plt.ylabel('sequence retrieval score',fontdict={'fontsize':20})
plt.axis([0, 50, 0, 1])
plt.legend(loc='upper right');
plt.show()
In [2]:
data = np.load('/Users/pablo/Desktop/data.npy')[()]
In [15]:
key = (5,15)
test_overlaps = data[key][54]
pt.plot_overlaps(test_overlaps[:,:])
Out[15]:
In [16]:
ev.my_score(key, [data[key][54]], cutoff = 0.75, time_retrieved = 15)
Out[16]:
In [1]:
# General libraries
import os
import pickle
import numpy as np
# Plot, in nb, only when .show() is called
import matplotlib.pyplot as plt
%matplotlib notebook
plt.ioff()
# Personal libraries
import tools.evaluation as ev
import tools.plot as pt
In [ ]:
path = '/home/sartori/Data/memexp'
# path = '/Users/pablo/Data/memexp'
dir_list = [ path + '/' + x for x in os.listdir(path) if x[0:10]=='KinCapCorr']
dir_list.sort()
In [ ]:
data = {}
for d in dir_list:
# Load data
overlaps = np.load(d + '/overlaps.npy')
# Create the key to store data
p = 16
c = int(d.split("_")[1][1:])
key = (c, p)
# Store the data
data.setdefault(key,[]).append(overlaps)
print d
In [ ]:
xyz = []
for key in data.iterkeys():
k = 15
p = key[1]
wowo = ev.my_score((k, p), data[key], cutoff = 0.75, time_retrieved = 15)
xyz.append( (k, key[0], wowo) )
XYZ = np.asarray(xyz)
XY = XYZ[:,1:]
XY[np.argsort(XY[:,0]),:]
In [4]:
Ek15 = np.array([[ 0.00000000e+00, 1.55333333e-01],
[ 1.00000000e+00, 6.66666667e-02],
[ 2.00000000e+00, 1.52666667e-01],
[ 3.00000000e+00, 1.64000000e-01],
[ 4.00000000e+00, 1.60000000e-01],
[ 5.00000000e+00, 1.76666667e-01],
[ 6.00000000e+00, 1.86666667e-01],
[ 7.00000000e+00, 1.85333333e-01],
[ 8.00000000e+00, 2.06000000e-01],
[ 9.00000000e+00, 2.34666667e-01],
[ 1.00000000e+01, 2.42000000e-01],
[ 1.10000000e+01, 2.57333333e-01],
[ 1.20000000e+01, 3.11333333e-01],
[ 1.30000000e+01, 3.15333333e-01],
[ 1.40000000e+01, 3.01333333e-01],
[ 1.50000000e+01, 3.48000000e-01],
[ 1.60000000e+01, 3.32000000e-01],
[ 1.70000000e+01, 3.24000000e-01],
[ 1.80000000e+01, 4.00000000e-01],
[ 1.90000000e+01, 4.26000000e-01],
[ 2.00000000e+01, 3.96000000e-01],
[ 2.10000000e+01, 3.52666667e-01],
[ 2.20000000e+01, 3.90666667e-01],
[ 2.30000000e+01, 4.13541667e-01],
[ 2.40000000e+01, 3.05333333e-01],
[ 2.50000000e+01, 2.25372549e-01],
[ 2.60000000e+01, 2.56588235e-01],
[ 2.70000000e+01, 1.68666667e-01],
[ 2.80000000e+01, 6.20000000e-02],
[ 2.90000000e+01, 4.54166667e-02],
[ 3.00000000e+01, 5.20000000e-02],
[ 3.10000000e+01, 3.58596491e-02],
[ 3.20000000e+01, 6.00000000e-03],
[ 3.30000000e+01, 2.33333333e-02],
[ 3.40000000e+01, 1.40000000e-02],
[ 3.50000000e+01, 1.05138889e-02],
[ 3.60000000e+01, 4.47619048e-03],
[ 3.70000000e+01, 6.66666667e-03],
[ 3.80000000e+01, 4.66666667e-03],
[ 3.90000000e+01, 3.11111111e-03]])
Ek7 = np.array([[ 0.00000000e+00, 2.68571429e-01],
[ 1.00000000e+00, 1.42857143e-01],
[ 2.00000000e+00, 3.04285714e-01],
[ 3.00000000e+00, 3.02857143e-01],
[ 4.00000000e+00, 3.61428571e-01],
[ 5.00000000e+00, 3.35714286e-01],
[ 6.00000000e+00, 3.31607143e-01],
[ 7.00000000e+00, 4.04285714e-01],
[ 8.00000000e+00, 4.05714286e-01],
[ 9.00000000e+00, 4.21607143e-01],
[ 1.00000000e+01, 4.82964286e-01],
[ 1.10000000e+01, 5.36964286e-01],
[ 1.20000000e+01, 5.48928571e-01],
[ 1.30000000e+01, 5.17142857e-01],
[ 1.40000000e+01, 5.23392857e-01],
[ 1.50000000e+01, 5.62777778e-01],
[ 1.60000000e+01, 5.96785714e-01],
[ 1.70000000e+01, 6.42500000e-01],
[ 1.80000000e+01, 6.36428571e-01],
[ 1.90000000e+01, 6.26785714e-01],
[ 2.00000000e+01, 6.23928571e-01],
[ 2.10000000e+01, 6.12039683e-01],
[ 2.20000000e+01, 4.96304945e-01],
[ 2.30000000e+01, 5.57634921e-01],
[ 2.40000000e+01, 4.75813492e-01],
[ 2.50000000e+01, 3.79519841e-01],
[ 2.60000000e+01, 3.07884921e-01],
[ 2.70000000e+01, 2.33285104e-01],
[ 2.80000000e+01, 1.86357143e-01],
[ 2.90000000e+01, 1.28000111e-01],
[ 3.00000000e+01, 1.36276718e-01],
[ 3.10000000e+01, 1.21539322e-01],
[ 3.20000000e+01, 5.11714674e-02],
[ 3.30000000e+01, 4.05880231e-02],
[ 3.40000000e+01, 3.30740093e-02],
[ 3.50000000e+01, 5.33261034e-02],
[ 3.60000000e+01, 6.31888061e-02],
[ 3.70000000e+01, 6.10113449e-02],
[ 3.80000000e+01, 3.93618355e-02],
[ 3.90000000e+01, 4.30572930e-02]])
Kk7 = np.array([[ 0.00000000e+00, 8.61299839e-01],
[ 1.00000000e+00, 1.42857143e-01],
[ 2.00000000e+00, 8.52142857e-01],
[ 3.00000000e+00, 7.95597763e-01],
[ 4.00000000e+00, 7.79185673e-01],
[ 5.00000000e+00, 7.71347403e-01],
[ 6.00000000e+00, 6.92350289e-01],
[ 7.00000000e+00, 6.85451673e-01],
[ 8.00000000e+00, 6.38240886e-01],
[ 9.00000000e+00, 6.32092768e-01],
[ 1.00000000e+01, 6.16261724e-01],
[ 1.10000000e+01, 6.00443307e-01],
[ 1.20000000e+01, 5.68620130e-01],
[ 1.30000000e+01, 5.65396825e-01],
[ 1.40000000e+01, 5.03704545e-01],
[ 1.50000000e+01, 5.00259740e-01],
[ 1.60000000e+01, 4.17658730e-01],
[ 1.70000000e+01, 4.05702381e-01],
[ 1.80000000e+01, 3.49523810e-01],
[ 1.90000000e+01, 3.37893218e-01],
[ 2.00000000e+01, 3.11447670e-01],
[ 2.10000000e+01, 2.43791209e-01],
[ 2.20000000e+01, 2.13593074e-01],
[ 2.30000000e+01, 1.64444444e-01],
[ 2.40000000e+01, 1.21648352e-01],
[ 2.50000000e+01, 1.05714286e-01],
[ 2.60000000e+01, 6.57142857e-02],
[ 2.70000000e+01, 5.15873016e-02],
[ 2.80000000e+01, 3.28571429e-02],
[ 2.90000000e+01, 1.57142857e-02],
[ 3.00000000e+01, 8.57142857e-03],
[ 3.10000000e+01, 1.00000000e-02],
[ 3.20000000e+01, 2.85714286e-03],
[ 3.30000000e+01, 1.42857143e-03],
[ 3.40000000e+01, 1.42857143e-03],
[ 3.50000000e+01, 9.09090909e-04],
[ 3.60000000e+01, 0.00000000e+00],
[ 3.70000000e+01, 0.00000000e+00],
[ 3.80000000e+01, 0.00000000e+00],
[ 3.90000000e+01, 0.00000000e+00]])
Kk15 = np.array([[ 0.00000000e+00, 4.44000000e-01],
[ 1.00000000e+00, 6.66666667e-02],
[ 2.00000000e+00, 3.95333333e-01],
[ 3.00000000e+00, 3.64666667e-01],
[ 4.00000000e+00, 3.33333333e-01],
[ 5.00000000e+00, 3.16000000e-01],
[ 6.00000000e+00, 3.03333333e-01],
[ 7.00000000e+00, 2.84666667e-01],
[ 8.00000000e+00, 2.65333333e-01],
[ 9.00000000e+00, 2.49333333e-01],
[ 1.00000000e+01, 2.24000000e-01],
[ 1.10000000e+01, 2.06666667e-01],
[ 1.20000000e+01, 2.16666667e-01],
[ 1.30000000e+01, 2.00000000e-01],
[ 1.40000000e+01, 1.76666667e-01],
[ 1.50000000e+01, 1.66000000e-01],
[ 1.60000000e+01, 1.34666667e-01],
[ 1.70000000e+01, 1.24000000e-01],
[ 1.80000000e+01, 1.00000000e-01],
[ 1.90000000e+01, 8.40000000e-02],
[ 2.00000000e+01, 7.73333333e-02],
[ 2.10000000e+01, 5.53333333e-02],
[ 2.20000000e+01, 4.93333333e-02],
[ 2.30000000e+01, 4.26666667e-02],
[ 2.40000000e+01, 2.73333333e-02],
[ 2.50000000e+01, 2.20000000e-02],
[ 2.60000000e+01, 1.80000000e-02],
[ 2.70000000e+01, 1.73333333e-02],
[ 2.80000000e+01, 8.00000000e-03],
[ 2.90000000e+01, 6.66666667e-03],
[ 3.00000000e+01, 1.33333333e-03],
[ 3.10000000e+01, 6.66666667e-04],
[ 3.20000000e+01, 6.66666667e-04],
[ 3.30000000e+01, 6.66666667e-04],
[ 3.40000000e+01, 0.00000000e+00],
[ 3.50000000e+01, 0.00000000e+00],
[ 3.60000000e+01, 0.00000000e+00],
[ 3.70000000e+01, 0.00000000e+00],
[ 3.80000000e+01, 0.00000000e+00],
[ 3.90000000e+01, 0.00000000e+00]])
In [5]:
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax1.scatter(Ek15[:,0], Ek15[:,1], s=35, c='b', marker="o", label='Cubic-interactions $p=16,\;k=15$')
ax1.scatter(Ek7[:,0], Ek7[:,1], s=35, c='g', marker="o", label='Cubic-interactions $p=16,\;k=7$')
ax1.scatter(Kk15[:,0], Kk15[:,1], s=35, c='lightblue', marker="o", label='Kanter-Sompolinsky $p=16,\;k=15$')
ax1.scatter(Kk7[:,0], Kk7[:,1], s=35, c='lightgreen', marker="o", label='Kanter-Sompolinsky $p=16,\;k=7$')
plt.xlabel('core size, $c$',fontdict={'fontsize':20})
plt.ylabel('sequence retrieval score',fontdict={'fontsize':20})
plt.axis([0, 50, 0, 1])
plt.legend(loc='upper right');
plt.show()
In [1]:
# General libraries
import os
import pickle
import numpy as np
# Plot, in nb, only when .show() is called
import matplotlib.pyplot as plt
%matplotlib notebook
plt.ioff()
# Personal libraries
import tools.evaluation as ev
import tools.plot as pt
In [ ]:
path = '/home/sartori/Data/memexp_sequence_length/En_p16_c20'
# path = '/Users/pablo/Data/memexp'
dir_list = [ path + '/' + x for x in os.listdir(path) if x[0:8]=='EnSeqLen']
dir_list.sort()
In [ ]:
data = {}
for d in dir_list:
# Load data
overlaps = np.load(d + '/overlaps.npy')
# Create the key to store data
p = 20
k = int(d.split("_")[5][1:])
key = (k, p)
# Store the data
data.setdefault(key,[]).append(overlaps)
print d
In [ ]:
with open(path+'/data.npy', 'wb') as f: np.save(f, data)
In [ ]:
xyz = []
for key in data.iterkeys():
k = key[0]
p = key[1]
wowo = ev.my_score((k, p), data[key], cutoff = 0.75, time_retrieved = 15)
xyz.append( (key[1], key[0], wowo) )
XYZ = np.asarray(xyz)
XY = XYZ[:,1:]
XY[np.argsort(XY[:,0]),:]
In [12]:
Kp16c0 = np.array([[ 1. , 1. ],
[ 2. , 0.98666667],
[ 3. , 0.971 ],
[ 4. , 0.971 ],
[ 5. , 0.91410281],
[ 6. , 0.88489569],
[ 7. , 0.83180519],
[ 8. , 0.7965348 ],
[ 9. , 0.71244444],
[ 10. , 0.66614286],
[ 11. , 0.57727273],
[ 12. , 0.56262821],
[ 13. , 0.50923077],
[ 14. , 0.47285714],
[ 15. , 0.41466667],
[ 16. , 0.39125 ],
[ 17. , 0.37529412],
[ 18. , 0.32166667]])
En16c0 = np.array([[ 1. , 0.985 ],
[ 2. , 0.715 ],
[ 3. , 0.58416667],
[ 4. , 0.4605 ],
[ 5. , 0.38833333],
[ 6. , 0.3052381 ],
[ 7. , 0.28285714],
[ 8. , 0.2825 ],
[ 9. , 0.22666667],
[ 10. , 0.216 ],
[ 11. , 0.18916667],
[ 12. , 0.1725 ],
[ 13. , 0.19538462],
[ 14. , 0.13428571],
[ 15. , 0.13466667],
[ 16. , 0.115 ],
[ 17. , 0.10764706],
[ 18. , 0.10722222]])
Kp16c20 = np.array([[ 1. , 0.91688095],
[ 2. , 0.81619048],
[ 3. , 0.7350487 ],
[ 4. , 0.56019391],
[ 5. , 0.48214835],
[ 6. , 0.32807113],
[ 7. , 0.2952381 ],
[ 8. , 0.26375 ],
[ 9. , 0.21444444],
[ 10. , 0.157 ],
[ 11. , 0.13181818],
[ 12. , 0.105 ],
[ 13. , 0.08923077],
[ 14. , 0.07071429]])
En16c20 = np.array([[ 1. , 0.91 ],
[ 2. , 0.77 ],
[ 3. , 0.81316667],
[ 4. , 0.7385 ],
[ 5. , 0.72269841],
[ 6. , 0.68404762],
[ 7. , 0.64701984],
[ 8. , 0.55072222],
[ 9. , 0.51480808],
[ 10. , 0.56598485],
[ 11. , 0.57325758],
[ 12. , 0.44965201],
[ 13. , 0.41461538],
[ 14. , 0.46438095]])
In [18]:
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax1.scatter(Kp16c0[:,0], Kp16c0[:,1], s=35, c='b', marker="o", label='Kanter-Sampolinsky $p=20,\;c=0$')
ax1.scatter(Kp16c20[:,0], Kp16c20[:,1], s=35, c='lightblue', marker="o", label='Kanter-Sampolinsky $p=16,\;c=20$')
ax1.scatter(En16c0[:,0], En16c0[:,1], s=35, c='g', marker="o", label='Cubic-interactions $p=20,\;c=0$')
ax1.scatter(En16c20[:,0], En16c20[:,1], s=35, c='lightgreen', marker="o", label='Cubic-interactions $p=16,\;c=20$')
plt.xlabel('sequence length, $k$',fontdict={'fontsize':20})
plt.ylabel('sequence retrieval score',fontdict={'fontsize':20})
plt.axis([0, 20, 0, 1])
plt.legend(loc='upper right');
plt.show()
In [ ]: