In [1]:

    

import matplotlib.pyplot as plt
import pylab as pl
import scipy.io as sio
import numpy as np
import cPickle as cp
import csv, glob, os


    

In [10]:

    

def data_stats(setname):
    data = np.zeros((16,))
    for fname in glob.glob("Slogs/" + str(setname) + "_*"):
        c_data = []
        rfile = open(fname, 'r')
        for row in rfile:
            if not row.startswith("#"):
                proc = row.strip().split(" ")
                c_data.append(proc[3])
        c_data = np.asarray(c_data, dtype=np.float32)
        best = np.min(c_data)
        subj_str = fname[14:16]
        if subj_str[-1] == "_": subj_str = fname[14]
        subject = int(subj_str)
        data[subject] = best
        
    data = np.asarray(data)
    return data


    

In [13]:

    

ssdec = (1-data_stats("SSALL")).reshape(-1,1)
np.mean(ssdec)


    

    
Out[13]:





0.6976878484711051

In [22]:

    

def CV_stats():
    data = np.zeros((16))
    for fname in glob.glob("TCV_SSMS" + "_*"):
        c_data = []
        rfile = open(fname, 'r')
        for row in rfile:
            if not row.startswith("#"):
                proc = row.strip().split(" ")
                c_data.append(proc[3])
        c_data = np.asarray(c_data, dtype=np.float32)
        best = np.min(c_data)
        subject = int(fname[15:])
        data[subject] = best
        
    data = np.asarray(data)
    return data


    

In [24]:

    

np.mean(CV_stats())


    

    
Out[24]:





0.361328125

In [24]:

    

def get_data(logname):
    c_data = []
    rfile = open(logname, 'r')
    for row in rfile:
        if not row.startswith("#"):
            proc = row.strip().split()
            c_data.append(proc)
    data = np.asarray(c_data)
    err_data = np.asarray(data[:,1:5], dtype=np.float32)
    return err_data


    

In [3]:

    

def proc_data(data_in):
    mu = data_in.mean(axis=0)
    std = data_in.std(axis=0)
    return np.concatenate([np.arange(len(mu)).reshape(-1,1), mu, std], axis=1)


    

In [4]:

    

%matplotlib inline


    

In [5]:

    

plot_info = []
data2 = proc_data(get_data("500l3", "0.80"))
plot_info.append({'X': data2[:,0], 'Y': data2[:,2], 'YSTD': data2[:,4], 'label': "0.80", 'color': 'b'})
data2 = proc_data(get_data("500l3", "0.90"))
plot_info.append({'X': data2[:,0], 'Y': data2[:,2], 'YSTD': data2[:,4], 'label': "0.90", 'color': 'g'})
data2 = proc_data(get_data("500l3", "0.98"))
plot_info.append({'X': data2[:,0], 'Y': data2[:,2], 'YSTD': data2[:,4], 'label': "0.98", 'color': 'r'})


    

In [27]:

    

data2 = get_data("test")
X, Y = data2[:,0], data2[:,3]
plt.plot(np.log(1+X), Y)

plt.legend()
plt.axhline(0.09, color='k', linestyle='--')
plt.xlabel('Epoch')
plt.ylabel('CV classification score (% correct)')
plt.show()


    

In [ ]:

    

def gen_plot(X, Y, YSTD, item, show=False):
    plt.plot(X, Y, label=item)
    plt.axhline(0.50, color='k', linestyle='--',
                label="Chance level")
    #plt.axvline(0, color='r', label='stim onset')
    plt.legend()
    hyp_limits = (Y - YSTD, Y + YSTD)
    plt.fill_between(X, hyp_limits[0],
                     y2=hyp_limits[1], color='b', alpha=0.5)
    plt.xlabel('Epoch')
    plt.ylabel('CV classification score (% correct)')
    #plt.ylim([30, 100])
    plt.title('Momentum parameter tests')
    if show:
        plt.show()
    else:
        plt.savefig("plots/"+item+"plot.png")
        plt.clf()


    

In [ ]:

    

#items = ["0.80", "0.90", "0.91", "0.92", "0.93", "0.94", "0.95", "0.98"]
#items = ["0.020", "0.010", "0.005", "0.002"]
#items = ["20_", "40_", "60_", "80_", "100_"]
#items = ["100_", "200_", "300_", "400_", "500_"]
items = ["0_", "2_", "5_", "7_", "8_", "12_"]
for item in items:
    data1 = get_data("rawMCV", item)
    data2 = proc_data(data1)
    #np.savetxt("plog_"+item, data2, fmt='%.5e')
    gen_plot(data2[:,0], data2[:,2], data2[:,4], item)