return ((y - y_pred)**2).mean()
def accuracy(y, y_pred):
pred = (y_pred >= 0.5).reshape(-1)
a = (pred == y).mean()
p = (y[pred == 1]).mean()
r = (pred[y == 1]).mean()
return a, p, r
lr = 0.3
inp = 28 * 28
hidden = 500
epoch = 20
batch_size = 32
n = nn(inp, hidden, lr)
for i in range(epoch):
for X, y in iiter(trainX, trainY, batch_size):
n.feedforward(X)
n.Loss(y)
n.backprop()
n.feedforward(trainX)
print("train loss\t", error(trainY, n.output))
n.feedforward(testX)
print("test loss\t", error(testY, n.output))
print("measures: \t", accuracy(testY, n.output))
from NN import NeuralNetwork as nn import numpy as np n = nn(3, 2, 0.8) n.weights1 = np.array([[0.35, 0], [0.15, -0.1], [-0.2, 0.2]]) n.weights2 = np.array([[0.4], [0.25]]) n.bias1 = np.array([0, 0]).reshape(1, 2) n.bias2 = np.array([0]).reshape(1, 1) n.lr = 0.8 x = np.array([0.5, 0.3, 0.9]).reshape(1, 3) n.feedforward(x) n.Loss(0.8) n.backprop() print(n.weights2) print(n.weights1)
import pandas as pd
from DMC import dmc
from KNN import knn
from NN import nn
data = pd.read_csv("iris.csv")
testSet = [[7.2, 3.6, 5.1, 2.5]]
test = pd.DataFrame(testSet)
k = 5
result1, neighbor1 = nn(data, test)
result2, neighbor2 = knn(data, test, k)
result3, neighbor3 = dmc(data, test)
print("\nResultados: ")
print("NN\n\tResults: {} - Vizinho: {}".format(result1, neighbor1))
print("KNN\n\tResults: {} - Vizinho: {}".format(result2, neighbor2))
print("DMC\n\tResults: {} - Vizinho(Centroide): {}".format(result3, neighbor3))