darden1

from sklearn.neural_network import MLPClassifier
from sklearn.metrics import accuracy_score

# 教師データを通常の表記に戻す(sikit-learnの多クラスロジスティック回帰ではonehotにしなくてよい)
y_train = np.array([np.argmax(yi) for yi in Y_train])
y_test = np.array([np.argmax(yi) for yi in Y_test])

batch_size = int(len(X_train)*0.2)  # ミニバッチサイズ
epochs = 100  # エポック数
mu = 0.05 # 学習率

darden1

import matplotlib.pyplot as plt
from sklearn import datasets
from sklearn.preprocessing import OneHotEncoder
from sklearn.preprocessing import StandardScaler
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split

# アヤメデータセット
iris = datasets.load_iris()
X = iris.data

darden1

class MultiLayerPerceptron():

    def __init__(self, hidden_layer_sizes=(10,), activation=“relu”, random_state=0):
        self.init_state = True   # 重みの初期化判定フラグ
        self.dense = None        # 全結合インスタンスリスト
        self.act_func = None     # 活性化関数インスタンスリスト
        self.n_units = None      #　各レイヤーのユニット数リスト
        self.loss = None         # トレーニングデータのLoss
        self.val_loss = None     # テストデータのLoss
        self.acc = None          # トレーニングデータの正答率

darden1

class Dense:
    def __init__(self, units, input_dim, kernel_initializer='he_normal', bias_initializer='zeros'):
        self.units = units
        self.input_dim = input_dim
        self.kernel_initializer = kernel_initializer
        self.bias_initializer = bias_initializer
        self.W = None
        self.b = None
        self.dW = None
        self.db = None

darden1

class SoftMax():
    def __init__(self, name="softmax"):
        self.name = name

    def forward_prop(self, Z):
        self.Z = Z
        return self.softmax(Z)

    def back_prop(self, Z, Y):
        return self.grad_softmax_with_loss(Z, Y)

darden1

# -*- coding: utf-8 -*-
import numpy as np

class Activation():
    def __init__(self, name="sigmoid"):
        self.name = name

    def forward_prop(self, Z):
        if self.name=="sigmoid":
            return self.sigmoid(Z)

darden1

import numpy as np

n = 3
m = 3

B = np.random.rand(n,m)

Y = np.zeros_like(B)
for i in range(Y.shape[0]):
    Y[i, np.random.choice(np.arange(Y.shape[1]), 1)] = 1

darden1

# -*- coding: utf-8 -*-
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score

# 教師データを通常の表記に戻す(sikit-learnの多クラスロジスティック回帰ではonehotにしなくてよい)
y_train = np.array([np.argmax(yi) for yi in Y_train])
y_test = np.array([np.argmax(yi) for yi in Y_test])

# 正則化項:1/C/2*W^2 →C = ∞で正則化0
clf_sk = LogisticRegression(C=1e10, max_iter=500, solver="sag", tol=1e-10)

darden1

# -*- coding: utf-8 -*-
import matplotlib.pyplot as plt
from sklearn import datasets
from sklearn.preprocessing import OneHotEncoder
from sklearn.preprocessing import StandardScaler
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split


# アヤメデータセット

darden1

# -*- coding: utf-8 -*-
import numpy as np

class MultiClassLogisticRegression():

    def __init__(self):
        self.init_state = True        # 重みの初期化判定フラグ
        self.W = None                 # 重み
        self.b = None                 # 閾値
        self.loss = np.array([])      # トレーニングデータのLoss