Esempio n. 1
0
def example(nfeature,m,case,nclass=2):
	X = 4*np.random.rand(nfeature,m)-2
	if case == "linear":
		Y = X[0,:] + X[1,:] - 0.25
	elif case == "quadratic":
		Y = X[1,:] - np.square(X[0,:]) + 1.5
	elif case == "cubic":
		Y = X[1,:] - np.power(X[0,:],3) - 2*np.power(X[0,:],2)+ 1.5
	elif case == "disk":
		Y = np.square(X[0,:])+np.square(X[1,:])-1
	elif case == "ring":
		Y = 1.25*np.sqrt(np.square(X[0,:])+np.square(X[1,:]))
		Y = np.fmod(Y,nclass)
	elif case == "band":
		Y = X[0,:] + X[1,:] 
		Y = np.fmod(Y,nclass)
	Y = np.maximum(Y,0.0)
	Y = np.round(Y)
	Y = np.minimum(Y,nclass-1)
	Y = np.expand_dims(Y,axis=0)
	return X,Y

if __name__ == "__main__":
	nfeature = 2
	m = 2000
	case = "ring"
	nclass = 2
	X,Y = example(nfeature,m,case,nclass)
	plot_results.plot_results_data((X,Y),nclass)
	plt.show()
Esempio n. 2
0
def example(nfeature, m, case, nclass=2):
    X = 4 * np.random.rand(nfeature, m) - 2
    if case == "linear":
        Y = X[0, :] + X[1, :] - 0.25
    elif case == "quadratic":
        Y = X[1, :] - np.square(X[0, :]) + 1.5
    elif case == "cubic":
        Y = X[1, :] - np.power(X[0, :], 3) - 2 * np.power(X[0, :], 2) + 1.5
    elif case == "disk":
        Y = np.square(X[0, :]) + np.square(X[1, :]) - 1
    elif case == "ring":
        Y = 1.25 * np.sqrt(np.square(X[0, :]) + np.square(X[1, :]))
        Y = np.fmod(Y, nclass)
    elif case == "band":
        Y = X[0, :] + X[1, :]
        Y = np.fmod(Y, nclass)
    Y = np.maximum(Y, 0.0)
    Y = np.round(Y)
    Y = np.minimum(Y, nclass - 1)
    Y = np.expand_dims(Y, axis=0)
    return X, Y


if __name__ == "__main__":
    nfeature = 2
    m = 2000
    case = "ring"
    nclass = 2
    X, Y = example(nfeature, m, case, nclass)
    plot_results.plot_results_data(X, Y, nclass)
    plt.show()
    if noise:
        Y = Y + 0.3 * np.random.randn(X.shape[1])
    Y = np.maximum(Y, 0.0)
    Y = np.round(Y)
    Y = np.minimum(Y, nclass - 1)
    Y = np.expand_dims(Y, axis=0)
    # train and validation/test sets
    if validpercent == 0:
        return X, Y
    else:
        return X[:, :m], Y[:, 0:m], X[:, m:], Y[:, m:]


if __name__ == "__main__":
    nfeature = 2
    m = 1000
    case = "quadratic"
    nclass = 2
    noise = False
    validpercent = 0.0
    if validpercent > 0:
        Xtrain, Ytrain, Xvalid, Yvalid = example(nfeature, m, case, nclass,
                                                 noise, validpercent)
        plot_results.plot_results_data((Xtrain, Ytrain),
                                       nclass,
                                       validation_data=(Xvalid, Yvalid))
    else:
        Xtrain, Ytrain = example(nfeature, m, case, nclass, noise,
                                 validpercent)
        plot_results.plot_results_data((Xtrain, Ytrain), nclass)
    plt.show()