Yann LeCun都推荐的深度学习资料合集!

传统机器学习
  • 感知器

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/basic-ml/perceptron.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/basic-ml/perceptron.ipynb

  • 逻辑回归

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/basic-ml/logistic-regression.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/basic-ml/logistic-regression.ipynb

  • Softmax 回归(多项逻辑回归)

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/basic-ml/softmax-regression.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/basic-ml/softmax-regression.ipynb

多层感知器

  • 多层感知器

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mlp/mlp-basic.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mlp/mlp-basic.ipynb

  • 具有 Dropout 的多层感知器

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mlp/mlp-dropout.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mlp/mlp-dropout.ipynb

  • 具有批量归一化的多层感知器

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mlp/mlp-batchnorm.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mlp/mlp-batchnorm.ipynb

  • 具有从头开始反向传播的多层感知器

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mlp/mlp-lowlevel.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mlp/mlp-fromscratch__sigmoid-mse.ipynb

卷积神经网络基本

  • 卷积神经网络

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/cnn/cnn-basic.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-basic.ipynb

  • 具有 He 初始化的卷积神经网络

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-he-init.ipynb

概念

  • 用等效卷积层替换全连接

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/fc-to-conv.ipynb

全卷积

  • 全卷积神经网络

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-allconv.ipynb

AlexNet

  • CIFAR-10 上的 AlexNet

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-alexnet-cifar10.ipynb

VGG

  • 卷积神经网络 VGG-16

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/cnn/cnn-vgg16.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-vgg16.ipynb

  • 在 CelebA 上训练的 VGG-16 性别分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-vgg16-celeba.ipynb

  • 卷积神经网络 VGG-19

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-vgg19.ipynb

ResNet

  • ResNet 与残差块

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/resnet-ex-1.ipynb

  • 在 MNIST 上训练的 ResNet-18 数字分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet18-mnist.ipynb

  • 在 CelebA 上训练的 ResNet-18 性别分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet18-celeba-dataparallel.ipynb

  • 在 MNIST 上训练的 ResNet-34 数字分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet34-mnist.ipynb

  • 在 CelebA 上训练的 ResNet-34 性别分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet34-celeba-dataparallel.ipynb

  • 在 MNIST 上训练的 ResNet-50 数字分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet50-mnist.ipynb

  • 在 CelebA 上训练的 ResNet-50 性别分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet50-celeba-dataparallel.ipynb

  • 在 CelebA 上训练的 ResNet-101 性别分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet101-celeba.ipynb

  • 在 CIFAR-10 上训练的 ResNet-101

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet101-cifar10.ipynb

  • 在 CelebA 上训练的 ResNet-152 性别分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet152-celeba.ipynb

网络中的网络

  • CIFAR-10 分类器网络中的网络

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/nin-cifar10.ipynb

度量学习

  • 具有多层感知器的孪生网络

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/metric/siamese-1.ipynb

自编码器全连接自编码器

  • 自编码器

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/autoencoder/ae-basic.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-basic.ipynb

卷积自编码器

  • 具有解卷积 / 转置卷积的卷积自编码器

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/autoencoder/ae-deconv.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-deconv.ipynb

  • 具有解卷积(不具有池化操作)的卷积自编码器

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/autoencoder/ae-conv-nneighbor.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-conv-nneighbor.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-deconv-nopool.ipynb

    • 具有最近邻插值的卷积自编码器

  • 在 CelebA 上训练的具有最近邻插值的卷积自编码器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-conv-nneighbor-celeba.ipynb

  • 在 Quickdraw 上训练的具有最近邻插值的卷积自编码器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-conv-nneighbor-quickdraw-1.ipynb

变分自编码器

  • 变分自编码器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-conv-var.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-var.ipynb

    • 卷积变分自编码器

条件变分自编码器

  • 条件变分自编码器(具有重构损失中的标签)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-cvae.ipynb

  • 条件变分自编码器(不具有重构损失中的标签)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-cvae_no-out-concat.ipynb

  • 卷积条件变分自编码器(具有重构损失中的标签)

    • PYTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-cnn-cvae.ipynb

  • 卷积条件变分自编码器(不具有重构损失中的标签)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/autoencoder/ae-cnn-cvae_no-out-concat.ipynb

生成对抗网络

  • MNIST 上的全连接生成对抗网络

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/gan/gan.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/gan/gan.ipynb

  • MNIST 上的卷积生成对抗网络

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/gan/gan-conv.ipynb

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/gan/gan-conv.ipynb

  • MNIST 上具有标签平滑的卷积生成对抗网络

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/gan/gan-conv-smoothing.ipynb

递归神经网络多对一:情感分析 / 分类

  • 简单的单层递归神经网络(IMDB)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/rnn_simple_imdb.ipynb

  • 打包序列以忽略填充字符的简单单层递归神经网络(IMDB)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/rnn_simple_packed_imdb.ipynb

  • 具有长短期记忆网络单元的递归神经网络(IMDB)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/rnn_lstm_packed_imdb.ipynb

  • 具有长短期记忆网络单元和经预训练的 GloVe 词向量的递归神经网络(IMDB)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/rnn_lstm_packed_own_csv_imdb.ipynb

  • 具有长短期记忆网络单元和 CSV 格式的自有数据集的递归神经网络(IMDB)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/rnn_lstm_packed_own_csv_imdb.ipynb

  • 具有 GRU 单元的递归神经网络(IMDB)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/rnn_gru_packed_imdb.ipynb

  • 多层双向递归神经网络(IMDB)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/rnn_gru_packed_imdb.ipynb

多对多 / 序列到序列

  • 为生成新文本(Charles Dickens)的简单字符递归神经网络

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/char_rnn-charlesdickens.ipynb

序数回归

  • 序数回归卷积神经网络——CORAL w. ResNet34 on AFAD-Lite

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/ordinal/ordinal-cnn-coral-afadlite.ipynb

  • 序数回归卷积神经网络——Niu et al. 2016 w. ResNet34 on AFAD-Lite

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/ordinal/ordinal-cnn-niu-afadlite.ipynb

  • 序数回归卷积神经网络——Beckham and Pal 2016 w. ResNet34 on AFAD-Lite

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/ordinal/ordinal-cnn-beckham2016-afadlite.ipynb

要诀与技巧

  • 周期学习率

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/tricks/cyclical-learning-rate.ipynb

PyTorch 工作流和机制自定义数据集

  • 为自定义数据集使用 PyTorch 数据集加载实用程序——CSV 文件转换为 HDF5

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/custom-data-loader-csv.ipynb

  • 为自定义数据集使用 PyTorch 数据集加载使用程序——来自 CelebA 的面部图像

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/custom-data-loader-celeba.ipynb

  • 为自定义数据集使用 PyTorch 数据集加载使用程序——来自 Quickdraw 的图像

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/custom-data-loader-quickdraw.ipynb

  • 为自定义数据集使用 PyTorch 数据集加载使用程序——来自街景门牌号(SVHN)数据集的图像

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/custom-data-loader-svhn.ipynb

  • 为自定义数据集使用 PyTorch 数据集加载使用程序——亚洲人面部数据集(AFAD)

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/custom-data-loader-afad.ipynb

  • 为自定义数据集使用 PyTorch 数据集加载使用程序——历史彩色图像

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/custom-data-loader_dating-historical-color-images.ipynb

训练与预处理

  • 生成验证集拆分

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/validation-splits.ipynb

  • 具有固定内存的数据加载

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet34-cifar10-pinmem.ipynb

  • 图像标准化

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-standardized.ipynb

  • 图像转换示例

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/torchvision-transform-examples.ipynb

  • 具有自己的文本文件的 Char-RNN

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/char_rnn-charlesdickens.ipynb

  • 具有自己的 CSV 文件的情感分类递归神经网络

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/rnn/rnn_lstm_packed_own_csv_imdb.ipynb

并行计算

  • 使用数据并行的多 GPU——VGG-16 CelebA 上的性别分类器

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-vgg16-celeba-data-parallel.ipynb

其他

  • 顺序 API 和钩子

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/mlp-sequential.ipynb

  • 层内权重共享

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/cnn-weight-sharing.ipynb

  • 只使用 Matplotlib 在 Jupyter Notebook 绘制实时训练性能

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/plot-jupyter-matplotlib.ipynb

Autograd

  • 在 PyTorch 中获取中间变量的梯度

    • PyTorch:https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/mechanics/manual-gradients.ipynb

TensorFlow 工作流和机制自定义数据集

  • 为 Mini-batch 训练使用 NumPy NPZ Archives 进行组块图像数据集

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mechanics/image-data-chunking-npz.ipynb

  • 为 Mini-batch 使用 HDF5 进行存储图像数据集

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mechanics/image-data-chunking-hdf5.ipynb

  • 使用输入管道从 TFRecords 文件读取数据

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mechanics/tfrecords.ipynb

  • 使用 Queue Runners 从硬盘直接馈入图像

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mechanics/file-queues.ipynb

  • 使用 TensorFlow 的数据集 API

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mechanics/dataset-api.ipynb

训练和预处理

  • 保存和加载训练过的模型——从 TensorFlow 检查点文件和 NumPy NPZ Archives

    • TensorFlow:https://github.com/rasbt/deeplearning-models/blob/master/tensorflow1_ipynb/mechanics/saving-and-reloading-models.ipynb


上一篇:在jupyter notebook中调用(import).ipynb .py文件


下一篇:如何最简单、通俗地理解Python的文件?