【深層学習入門】Kerasによる回帰ニューラルネットワークの実装

from tensorflow.keras.datasets import boston_housing

(train_data, train_labels), (test_data, test_labels) = boston_housing.load_data()

print(train_data.shape)
print(train_labels.shape)
print(test_data.shape)
print(test_labels.shape)

## 出力
(404, 13)
(404,)
(102, 13)
(102,)

.
├── requirements.txt
└── app/
    ├── logs/
    ├── model.py
    ├── params.py
    ├── predict.py
    ├── preprocessing.py
    ├── train.py
    └── utils.py

numpy==12.2.4
pandas==1.4.2
matplotlib==3.5.2
seaborn==0.11.2
scikit-learn==1.1.1
nptyping==2.1.1
tensorflow==2.9.1
# tensorflow-macos==2.9.2  # M1macの場合はこちらを使用する
pydot==1.4.2
graphviz==0.20
rich==12.5.1

from datetime import datetime
from pathlib import Path

INPUT_SIZE = 13
HIDDEN1_SIZE = 64
HIDDEN2_SIZE = 64
OUTPUT_SIZE = 1

EPOCHS = 500
VALIDATION_SPLIT = 0.2

LOG_DIR = Path('logs/fit') / datetime.now().strftime("%Y%m%d-%H%M%S")
MODEL_FILE_PATH = Path('model.h5')

import tensorflow as tf
from keras.api._v2 import keras
from keras.layers import Input, Dense
from keras.models import Model

import params


class DNNModel:
    def __init__(
            self,
            input_size: int = params.INPUT_SIZE,
            hidden1_size: int = params.HIDDEN1_SIZE,
            hidden2_size: int = params.HIDDEN2_SIZE,
            output_size: int = params.OUTPUT_SIZE) -> None:
        self.input: Input = Input(shape=(input_size,), name='input')
        self.hidden1: Dense = Dense(hidden1_size, activation='relu', name='hidden1')
        self.hidden2: Dense = Dense(hidden2_size, activation='relu', name='hidden2')
        self.output: Dense = Dense(output_size, name='output')

    def build(self) -> Model:
        input = self.input
        x = self.hidden1(input)
        x = self.hidden2(x)
        output = self.output(x)
        return Model(inputs=input, outputs=output)

import numpy as np
from nptyping import NDArray, Shape, Float, Int
from sklearn.preprocessing import StandardScaler

from utils import save_pkl, read_pkl


Features = NDArray[Shape['Sample, Feature'], Float]

def preprocess_dataset(dataset: Features, is_training: bool) -> Features:
    if is_training:
        sc = StandardScaler()
        dataset = sc.fit_transform(dataset)
        save_pkl(sc, path='scaler.pkl')
        return dataset

    sc = read_pkl(path='scaler.pkl')
    dataset = sc.transform(dataset)
    return dataset

import pickle


def save_pkl(obj, path):
    with open(path, 'wb') as f:
        pickle.dump(obj, f)

def read_pkl(path):
    with open(path, 'rb') as f:
        return pickle.load(f)
        

import tensorflow as tf
from keras.api._v2 import keras
from keras.models import Model
from keras.datasets import boston_housing
from keras.callbacks import EarlyStopping, ModelCheckpoint, TensorBoard
from keras.utils import plot_model

import params
from preprocessing import preprocess_dataset
from model import DNNModel


def main():
    (train_data, train_labels), (test_data, test_labels) = boston_housing.load_data()
    train_data, train_labels = preprocess_dataset(dataset=train_data, labels=train_labels)

    model: Model = DNNModel().build()

    model.compile(
        optimizer='adam',
        loss='mse',
        metrics=['mae'])
    model.summary()
    plot_model(model, to_file='model.pdf', show_shapes=True)

    callbacks = [
        EarlyStopping(patience=20),
        ModelCheckpoint(filepath=params.MODEL_FILE_PATH, save_best_only=True),
        TensorBoard(log_dir=params.LOG_DIR)]

    model.fit(
        x=train_data,
        y=train_labels,
        epochs=params.EPOCHS,
        validation_split=params.VALIDATION_SPLIT,
        callbacks=callbacks)
    
if __name__ == '__main__':
    main()

import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
from keras.api._v2 import keras
from keras.datasets import boston_housing
from rich import print

import params
from preprocessing import preprocess_dataset


def predict(dataset):
    model = keras.models.load_model(params.MODEL_FILE_PATH)
    pred = model.predict(dataset).flatten()
    return pred


if __name__ == '__main__':
    (train_data, train_labels), (test_data, test_labels) = boston_housing.load_data()
    test_data, test_labels = test_data[:10], test_labels[:10]  # 10データ分だけ推論をおこなう
    test_data = preprocess_dataset(dataset=test_data, is_training=False)
    pred = predict(dataset=test_data)
    print(f'prediction: {np.round(pred, 1)}')
    print(f'labels: {test_labels}')

$ python train.py

## 出力
Model: "model"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 input (InputLayer)          [(None, 13)]              0         
                                                                 
 hidden1 (Dense)             (None, 64)                896       
                                                                 
 hidden2 (Dense)             (None, 64)                4160      
                                                                 
 output (Dense)              (None, 1)                 65        
                                                                 
=================================================================
Total params: 5,121
Trainable params: 5,121
Non-trainable params: 0
_________________________________________________________________
Epoch 1/500
11/11 [==============================] - 0s 10ms/step - loss: 556.1224 - mae: 21.7004 - val_loss: 638.2590 - val_mae: 23.3441
Epoch 2/500
11/11 [==============================] - 0s 2ms/step - loss: 513.5609 - mae: 20.6846 - val_loss: 587.2802 - val_mae: 22.2735
.
.
.
Epoch 121/500
11/11 [==============================] - 0s 5ms/step - loss: 6.7154 - mae: 1.8254 - val_loss: 12.1797 - val_mae: 2.5738
Epoch 122/500
11/11 [==============================] - 0s 5ms/step - loss: 6.7295 - mae: 1.8319 - val_loss: 12.2754 - val_mae: 2.5554

$ tensorboard --logdir ./logs/fit

## 出力
Serving TensorBoard on localhost; to expose to the network, use a proxy or pass --bind_all
TensorBoard 2.9.1 at http://localhost:6006/ (Press CTRL+C to quit)

$ python predict.py

## 出力
prediction: [ 8.3 18.  21.1 34.6 26.  19.1 26.  21.6 19.6 22.3]
labels:     [ 7.2 18.8 19.  27.  22.2 24.5 31.2 22.9 20.5 23.2]