Articles by Yoshiyuki Kobayashi

Tutorial. Designing and training a two-layer neural network

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

1 Creating a new project Start Neural Network Console, and click + New Project to create a new project. 2 Assigning the datasets On the DATASET tab, click the dataset assignment buttons, and assign small_mnist_4or9_training.csv to Training   Click Validation and the dataset assignment buttons, and assign small_mnist_4or9_test.csv to Validation.   3 Designing the two-layer neural network Under “Component” on the left side of the EDIT tab, double-click on Input, Affine, Tanh, Affine, Sigmoid, and BinaryCrossEntropy, one by one, in…

Read More

Setup window

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

  To open the setup window, click the Setup button in the top right of the window.   1 Using the engine setup window   1.1 Switching between CPU execution and GPU execution Click CPU or GPU next to Processor Type.   1.2 Selecting the GPU to use on a multiple GPU environment Use GPU Device Index to specify the index (0 to 3) of the GPU to be used.   1.3 Changing the location of the related libraries that…

Read More

Project window. Other functions available on the project window

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

1 Creating a training result file of the network structure being edited without training Right-click the Run Training button to open a shortcut menu, and click Create Training Result.   Reference This function can be used to output training results of computation graph without parameters, output of new training results of edited network after being loaded after loading a network with the Open in EDIT Tab with Weight function and editing it, confirmation of initial values of a neural network…

Read More

Project window. Using the EVALUATION tab

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

  The EVALUATION tab is used to perform computation on a validation dataset using a trained neural network and evaluate the results.   1 Evaluating a trained neural network On the training result list, click the training result you want to evaluate. Click the Run Evaluation button. Alternatively, press F6 on the keyboard.   2 Viewing the evaluation progress View the evaluation progress information and log.   DATA in the evaluation progress information area shows the number of data samples…

Read More

Project window. Using the TRAINING tab

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

  The TRAINING tab is used to view the progress of the current training and compare several training results.   1 Viewing the training progress View the training progress information, graph monitor, and log.   EPOCH in the training progress information area indicates how many epochs have been completed, out of the total number of optimization epochs (Max Epoch) for a complete training as specified on the CONFIG tab.   Elapsed, Remaining, and Total in the training progress information area…

Read More

Project window. Using Executor on the CONFIG tab

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

  1 Specifying the name of the network used for evaluation performed with the Run Evaluation button Set Network to the name of the network created on the EDIT tab.   Reference By default, the network name is set to the MainRuntime network, which is automatically generated from the Main network created on the EDIT tab.   2 Specifying the name of the dataset used for evaluation performed with the Run Evaluation button Set data to the name of the…

Read More

Project window. Using Monitor on the CONFIG tab

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

  1 Specifying the name of the network used for evaluation during training Set Network to the name of the network created on the EDIT tab.   Reference The train error and the validation error on the learning curve are plotted according to the Monitor settings on the CONFIG tab. By default, the network name is set to the MainValidation network, which is automatically generated from the Main network created on the EDIT tab.   2 Specifying the name of…

Read More

Project window. Using Optimizer on the CONFIG tab

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

  1 Specifying the name of the network used for optimization Set Network to the name of the network created on the EDIT tab.   2 Specifying the name of the dataset used for optimization Set data to the name of the dataset loaded on the DATASET tab.   3 Specifying the parameter update method From the Config list, select Optimizer. Select an updater from the following (“Adam” is used by default).   Updater Update expression Adadelta   $$g_t \leftarrow…

Read More

Project window. Using Global Config on the CONFIG tab

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

  The Global CONFIG tab is used to configure and view the settings of neural network training.   1 Editing the project comment Enter any comment you like using text in Project Description.   2 Specifying the number of epochs until the completion of training optimization Enter the number of epochs in Max Epoch. In order to simply output an untrained designed network with the parameter properties (e.g., Weight, Bias) specified in all layers that contain some parameters, set Max…

Read More

Project window. Using the DATASET tab

Friday, November 03, 2017

App

Posted by Yoshiyuki Kobayashi

The DATASET tab is used to specify and view the datasets that will be used in neural network training.   1 Specifying a CSV file containing training and validation datasets From the data list, select the dataset CSV file. As the name suggests, the default Training dataset is for training, and the Validation dataset is for validation. Click the Dataset Assignment button. In the dataset management window that appears, select the dataset.   2 Viewing the content of a dataset…

Read More