Deep Learning in R

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

Deep learning has a wide range of applications, from speech recognition, computer vision, to self-driving cars and mastering the game of Go. While the concept is intuitive, the implementation is often heuristic and tedious. We will take a stab at simplifying the process, and make the technology more accessible.

Example

We illustrate our approach with the venerable CIFAR-10 dataset. The following code snippet will download the data from its known location to a folder “data/cifar” inside the current working directory. It also creates a data source object for later use.

> library(rpud)
> ds <- rpudlCreateDataSource(
+      data.format="cifar10",
+      data.dir="data/cifar"
+ )

The training process requires us to pre-compute the mean value of the training data. The following will save the result to a file “cifar10-data.mean” inside the data source’s directory.

> rpudlFindTrainingDataMean(ds, filename="cifar10-data.mean")

We then create a model object using a script named cifar10_quick_train.prototxt. The script is written in Google Protocol Buffers format. There is a local copy of the script in the rpudl folder of the rpud installation directory, which is what we will use below. The script is based on the cifar10 example of the Caffe framework. If interested, you can also find the message type definition “rpudl_model.proto” in the same local folder.

> rpudl.path <- file.path(path.package(package="rpud"), "rpudl")
> model <- rpudl(
+       file.path(rpudl.path, "cifar10_quick_train.prototxt"),
+       data.source=ds
+ )

Now we are ready to train the model. The training process takes about 280 seconds on a GTX 660Ti. The result is a new model object with updated parameters. Here, we overwrite the previous model object for simplicity. Should we save the result in a separate model object, we can have a history of the trained models. It reaches about 72% accuracy after 4,000 iterations.

> system.time(model <- rpudlTrain(
+       model, batch=100, iter=4000))
    ....
training @iter 3800, cost 0.535023, accuracy 74%
training @iter 3900, cost 0.689107, accuracy 73%
training @iter 4000, cost 0.655459, accuracy 78%
training ends with average cost of 0.951724, and accuracy 71.78%
  user  system elapsed
210.04   70.17  279.78

We then resume the training with a smaller learning rate. It takes about 72 seconds for another 1,000 iterations, and improves the accuracy to 75%.

> system.time(model <- rpudlTrain(
+       model, batch=100, iter=1000, learning.rate=0.0001)
+       )
    ....
training @iter 4800, cost 0.416993, accuracy 72%
training @iter 4900, cost 0.43716, accuracy 79%
training @iter 5000, cost 0.456889, accuracy 77%
training ends with average cost of 0.874285, and accuracy 75.43%
user  system elapsed
53.97   18.06   71.92

For demonstration purpose, we extract 12 samples from the testing data, and run prediction against the trained model.

> num <- 12
> obj <- rpudlGetTestingDataSamples(ds, c(1, num))
> res <- predict(model, obj$x)
> sum((obj$y) == res)
[1] 12

The model object keeps a history of model cost, which we can print out with the qplotmethod from the ggplot2 library.

> library(ggplot2)
> costs <- model$cost.log
> qplot(1:length(costs), costs, geom="smooth")

We can also have a picture of the weight parameters of a convolutional layer using a rpudlPlotLayerWeights method.

> weights <- rpudlGetLayerWeights(model, "conv1")
> rpudlPlotLayerWeights(weights)

Notes

• Currently, the rpudl methods are accessible even without an active rpudpluslicense.
• There are more examples to try out in the rpudl folder of the rpudinstallation directory.
• The supported data formats are CIFAR, MNIST and LMDB. To train with your own data, you have to export your data to an LMDB database according to the following message type:message Datum {
      required bytes data = 1;
      required uint32 label = 2;
  }
  
• The current object model consists of a simple processing pipeline. It does not support Long Short-Term Memory (LSTM) or Network-In-Network (NIN) models. We will address these limitations in upcoming releases.
  

References

• Yangqing Jia, Evan Shelhamer, et al: Caffe - A Deep Learning Frameworkhttp://caffe.berkeleyvision.org.
• Google Research: TensorFlow - An Open Source Software Library for Machine Intelligence https://www.tensorflow.org.
• Deep Learning: Theano - An Open Source Machine Learning Library forPython http://deeplearning.net/software/theano.
  

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏1 回帖

关键词：Learning earning Learn deep Earn technology following download computer learning

thank you for sharing

谢谢分享

谢谢分享

谢谢分享