1. 映画
《君の名は》
《僕の初恋をキミに捧ぐ》
《僕の彼女はサイボーグ》
《世界の中心で愛を叫ぶ》
《ナナ》
《ビリギャル》
《ウォーターボーイズ》
《とんび》
《おくりびと》
《武士の家計簿》
2. ドラマ
《1リットルの涙》
[……]
《君の名は》
《僕の初恋をキミに捧ぐ》
《僕の彼女はサイボーグ》
《世界の中心で愛を叫ぶ》
《ナナ》
《ビリギャル》
《ウォーターボーイズ》
《とんび》
《おくりびと》
《武士の家計簿》
《1リットルの涙》
[……]
Simple
1. umount the usb flash disk;
2. copy iso file to the device like follow command
debian@thinkpad:~/$ cp debian-12.10.0-amd64-DVD-1.iso /dev/sdx
the sdx is your usb device.
3. sync
debian@thinkpad:~/$ sync
Done!
reference link: https://www.debian.org/releases/bookworm//i386/ch04s03.en.h[……]
简单示例,一个用 CNN 识别 MNIST 手写数字的模型。数据是 60,000 张训练和 10,000 张测试的 28×28 灰度图,先归一化到 0-1,再加通道。模型用两层卷积(32 和 64 个滤波器)提取特征,两层池化缩小尺寸,再展平后用两个全连接层(128 和 10 个神经元)输出概率。训练 5 轮,用 adam 优化,损失是交叉熵。
import tensorflow as tf
import numpy as np
# 1. 加载MNIST 数据集
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.l[......]
展平与全连接得到特征值
import numpy as np
# 1、定义输入图片:4x4的图片,0表示白色,1表示黑色
image = np.array([
[0, 0, 1, 1],
[0, 0, 1, 1],
[1, 1, 0, 0],
[1, 1, 0, 0]
])
print('输入图片: \n', image)
# 2、卷积:用2x2滤波找对角特征
filter_simple = np.array([
[1, 0],
[0, 1]
])
print('滤波器: \n', filter_simple)
# 计算卷积结果
con[......]