def fit_and_display(X,Y, sample, deg):
N = len(X)
trainidx = np.random.choice(N, sample)
Xtrain = X[trainidx]
Ytrain = Y[trainidx]
plt.scatter(Xtrain, Ytrain)
plt.show()
#fit polynomial
# Xtrain_poly = ax.polyfy(Xtrain, deg)
X_poly = ax.polyfy(X, deg)
Xtrain_poly = X_poly[trainidx]
# w = fit(Xtrain_poly, Ytrain)
#Compute redisiduals
w, resid0 = lin(Xtrain_poly, Ytrain)
resid1 = r2(w, X_poly, Y)
print ("degree %d .: in-sample %.4f out sample %.4f" % (deg ,resid0, resid1))
#display the polynomial
# X_poly = ax.polyfy(X, deg)
Y_hat = X_poly.dot(w)
plt.plot(X,Y)
plt.plot(X,Y_hat)
plt.scatter(Xtrain, Ytrain)
plt.title("deg= %d" % deg)
plt.show()
def testLinearRegression1D(self):
df = pd.read_csv('../datasets/1d.csv', header=None)
X = df.loc[:, 0]
Y = df.loc[:, 1]
W, rsq = lin(X, Y)
self.assertAlmostEqual(W[0], 2.8644240756607786, None,
"W[0]*X + W[1] .: W[0] should ~ 2.8644", 0.0001)
self.assertAlmostEqual(W[1], 1.972612167484588, None,
"W[0]*X + W[1] .: W[1] should ~ 1.9726", 0.0001)
self.assertAlmostEqual(rsq, 0.99118382029778052, None, "r2 ~ 0.99118",
0.0001)
def testLinearRegression2D(self):
df = pd.read_csv('../datasets/2d.csv', header=None)
X = df.loc[:, [0, 1]]
Y = df.loc[:, 2]
W, rsq = lin(X, Y, fill=True)
self.assertAlmostEqual(W[0], 2.01666793, None,
"W[0] .: expected ~2.0166\ngot " + str(W[0]),
0.0001)
self.assertAlmostEqual(W[1], 2.96985048, None,
"W[1] .: expected ~2.96985\ngot " + str(W[1]),
0.0001)
self.assertAlmostEqual(W[2], 1.46191241, None,
"W[2] .: expected ~1.46191\ngot " + str(W[2]),
0.0001)
self.assertAlmostEqual(rsq, 0.99800406124757779, None,
"r2 .: ~0.99800\ngot " + str(rsq), 0.0001)
if __name__ == '__main__':
pass
N = 50
X = np.linspace(0, 10, N)
Y = 0.5 * X + np.random.randn(N)
Y[-1] += 30
Y[-2] += 30
plt.scatter(X, Y)
plt.show()
#Why??
X = np.vstack([np.ones(N), X]).T
# XX = np.concatenate((np.ones(N),np.array([N])))
wml, residml = lin(X, Y)
Yhat_ml = X.dot(wml)
print wml
print residml
l2 = 1000
wmap, residmap = linl2(X, Y, l2)
Yhat_map = X.dot(wmap)
plt.scatter(X[:, 1], Y)
plt.plot(X[:, 1], Yhat_ml, label=("ml r2=%.4f" % (residml)))
plt.plot(X[:, 1], Yhat_map, label=("map r2=%.4f" % (residmap)))
plt.legend()
plt.show()
'''
Created on May 10, 2017
@author: Varela
'''
import pandas as pd
import regress as rg
from aux import get_dataset
# import matplotlib.pyplot as plt
df = pd.read_csv(get_dataset('1d.csv'), header=None)
X = df.iloc[:, 0]
Y = df.iloc[:, 1]
W, r2 = rg.lin(X, Y)
print W
print r2
#plot the data to see what's look like
# plt.scatter(X, Y)
# plt.show()
# \[(.*)\] matches anything enclosed in brackets
# [^\d]+ matches anything which is not a decimal
Y = df.loc[:, 1].str.replace(r'\[(.*)\]', '')
Y = Y.str.replace(r'[^\d]+', '').astype('int')
X = df.loc[:, 2].str.replace(r'\[(.*)\]', '')
X = X.replace(r'[^\d]+', '').astype('int')
# plt.scatter(X, Y)
# plt.show()
#
# Y = np.log(Y)
# plt.scatter(X, Y)
# plt.show()
Y = np.log(Y)
W, r2 = lin(X, Y)
print W
print r2
#LAZY PROGRAMMER SOLUTION
# import re
# import numpy as np
# import matplotlib.pyplot as plt
#
# X = []
# Y = []
#
# non_decimal = re.compile(r'[^\d]+')
#
# for line in open('../src/datasets/moore.csv'):
# r = line.split('\t')