使用DBN学习MINIST数据集

最新推荐文章于 2025-11-25 01:27:38 发布

原创最新推荐文章于 2025-11-25 01:27:38 发布 · 2.4k 阅读

5 ·

CC 4.0 BY-SA版权

随感同时被 2 个专栏收录

75 篇文章

订阅专栏

Python 机器学习

29 篇文章

订阅专栏

本教程通过使用GitHub上的TensorFlow代码实现了一个深层信念网络（DBN）的预训练和微调过程。该过程包括从MNIST数据集中下载训练数据，并通过多个隐藏层进行逐层预训练。

部署运行你感兴趣的模型镜像

使用github上的tensorflow代码,数据下载和训练一条龙。

下载代码

#git clone https://github.com/xiaohu2015/DeepLearning_tutorials.git

运行

#python dbn.py

输出

Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.
Extracting MNIST_data/train-images-idx3-ubyte.gz
Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.
Extracting MNIST_data/train-labels-idx1-ubyte.gz
Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.
Extracting MNIST_data/t10k-images-idx3-ubyte.gz
Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.
Extracting MNIST_data/t10k-labels-idx1-ubyte.gz
2018-07-09 21:31:32.912218: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
Starting pretraining...

    Pretraing layer 0 Epoch 0 cost: 82.7562146932
    Pretraing layer 0 Epoch 1 cost: 71.4445186615
    Pretraing layer 0 Epoch 2 cost: 68.2782074252
    Pretraing layer 0 Epoch 3 cost: 66.6253933924
    Pretraing layer 0 Epoch 4 cost: 65.5671972968
    Pretraing layer 0 Epoch 5 cost: 64.8149403104
    Pretraing layer 0 Epoch 6 cost: 64.2372861134
    Pretraing layer 0 Epoch 7 cost: 63.7739256183
    Pretraing layer 0 Epoch 8 cost: 63.3849852024
    Pretraing layer 0 Epoch 9 cost: 63.0742568346
    Pretraing layer 1 Epoch 0 cost: 127.297717382
    Pretraing layer 1 Epoch 1 cost: 115.47870919
    Pretraing layer 1 Epoch 2 cost: 112.256735805
    Pretraing layer 1 Epoch 3 cost: 110.447800154
    Pretraing layer 1 Epoch 4 cost: 109.252376036
    Pretraing layer 1 Epoch 5 cost: 108.416442469
    Pretraing layer 1 Epoch 6 cost: 107.767917571
    Pretraing layer 1 Epoch 7 cost: 107.259497833
    Pretraing layer 1 Epoch 8 cost: 106.867456998
    Pretraing layer 1 Epoch 9 cost: 106.534078383
    Pretraing layer 2 Epoch 0 cost: 36.3754562274
    Pretraing layer 2 Epoch 1 cost: 32.9483315901
    Pretraing layer 2 Epoch 2 cost: 32.2201649111
    Pretraing layer 2 Epoch 3 cost: 31.8175771783
    Pretraing layer 2 Epoch 4 cost: 31.5295569212
    Pretraing layer 2 Epoch 5 cost: 31.3358987011
    Pretraing layer 2 Epoch 6 cost: 31.1839012215
    Pretraing layer 2 Epoch 7 cost: 31.056089396
    Pretraing layer 2 Epoch 8 cost: 30.9713914368
    Pretraing layer 2 Epoch 9 cost: 30.8702448134

The pretraining process ran for 10.1000482996 minutes

Start finetuning...

    Epoch 0 cost: 0.444020345374, validation accuacy: 0.935400009155
    Epoch 1 cost: 0.228727179088, validation accuacy: 0.947200000286
    Epoch 2 cost: 0.192814317359, validation accuacy: 0.956600010395
    Epoch 3 cost: 0.172022930289, validation accuacy: 0.958800017834
    Epoch 4 cost: 0.15778082135, validation accuacy: 0.962999999523
    Epoch 5 cost: 0.146718783832, validation accuacy: 0.963800013065
    Epoch 6 cost: 0.137771496312, validation accuacy: 0.965600013733
    Epoch 7 cost: 0.130382152335, validation accuacy: 0.9674000144
    Epoch 8 cost: 0.124080524492, validation accuacy: 0.969399988651
    Epoch 9 cost: 0.118702218932, validation accuacy: 0.970399975777

The finetuning process ran for 1.72220735153 minutes
[Finished in 780.0s]