def scenario3_1_3bis():
N = 100
X, T = generateDataSet(N, 1)
multipleLayerNN = NN2.MultipleLayer(batch_size=100,
nb_eboch=100,
lr=0.001,
hidden_layer_size=5)
Tbis = np.reshape(T, (len(T), 1))
WHistoryMNN, eHistoryMNN = multipleLayerNN.fit(X, Tbis)
graph.plotNNInformations("Multiple Layer NN", X, T, WHistoryMNN[-1],
eHistoryMNN)
graph.plotDecisionBoundaryAnim("Multiple Layer NN Anim", X, T, WHistoryMNN)
def scenario3_2_2():
N = 8
H = 3
X = np.ones((N, N)) * -1
indices_diagonal = np.diag_indices(N)
X[indices_diagonal] = 1
multipleLayerNN = NN2.MultipleLayer(hidden_layer_size=H,
batch_size=-1,
nb_eboch=10000,
lr=0.01)
WHistory, eHistory = multipleLayerNN.fit(X, X)
predict = multipleLayerNN.predict(X)
for i, p in enumerate(predict):
print("Encoder predicts : ")
print(p)
print("Good answer was : ")
print(X[i])
if np.array_equal(X[i], p):
print("GOOD!!!")
else:
print("FAIL!!!")
print("--------------------------")
graph.plotError("Encoder Learning Curve", eHistory)
def scenario3_2_1():
N = 100
V = 5
X = list(np.random.multivariate_normal([V, V], [[1, 0], [0, 1]],
N)) # Blue
X += list(np.random.multivariate_normal([-V, -V], [[1, 0], [0, 1]],
N)) # Blue
X += list(np.random.multivariate_normal([V, -V], [[1, 0], [0, 1]],
N)) # Red
X += list(np.random.multivariate_normal([-V, V], [[1, 0], [0, 1]],
N)) # Red
T = [1] * 2 * N + [-1] * 2 * N
p = np.random.permutation(len(X))
X, T = (np.array(X)[p]).T, np.array(T)[p]
Tbis = np.reshape(T, (len(T), 1))
multipleLayerNN = NN2.MultipleLayer(batch_size=-1,
nb_eboch=100,
lr=0.001,
hidden_layer_size=2)
WHistoryMNN, eHistoryMNN = multipleLayerNN.fit(X, Tbis)
graph.plotNNInformations("Multiple Layer NN", X, T, WHistoryMNN[-1],
eHistoryMNN)
graph.plotDecisionBoundaryAnim("Multiple Layer NN Anim", X, T, WHistoryMNN)
