1.0 Start Training 

Tap 'Training' to start training.

1.1 Training Data 

In order to train a neural network, it is first necessary to prepare training data. Training data file is CSV file format compliant with RFC 4180. For this app, it is available for both 'Regression' and 'Classification'. Refer to sample files for details on the training data.

Ex1(Regression): ax^2+b

Ex2(Classification): if((ax^2+b)>0,1,0)

1.2 Neural Network Structure

Neural network consists of an input layer, an output layer, and up to 3 hidden layers. If only one training data is selected for output training data and its values are only integers 1 or 0, it is recognized as 'Classification'. Otherwise, it is recognized as 'Regression'.

1.3 Hyperparameters

Some of optimization methods & overfitting remedies require hyperparameter settings. Hyperparameter settings can be modified by interrupting training. Refer to keywords for details.

1.4 Training

Long tap the graph to start/stop training.

2.0 Start Prediction

Tap 'Prediction' to start prediction.

2.1 Neural Network & Prediction

Prediction is available after training or neural network file import. Neural network file can be exported after training. Refer to sample files for details on the neural network file.

Ex1(Regression): ax^2+b

Ex2(Classification): if((ax^2+b)>0,1,0)

Character Set: 95 kinds of character sets are available.

Color Theme: 62 colors to choose from.

Font Size: 0 .. +3

Rotation: Auto / Landscape / Portrait

Vibration: Enable / Disable

Sample File List: Display / Not display

Training History: Record / Not record

Auto-save Frequency: 1,000 / 10,000 epochs

Parameter Initialization: Random / Fixed#01-99

Parameter Recording: Latest / Lowest cost

Graph: Draw Type: Draw all epochs. / Not draw mini-batch & dropout.

Graph: Epoch Axis Range: 100 / 1,000 epochs