Work with AI Studio

AI Studio allows you to create powerful deep learning AI models without adding code, using a simple point-and-click interface. Whether you’re a deep learning novice or busy data scientist, this straightforward tool massively accelerates AI adoption in your organization

Overview

The AI Studio is the design center where you can train your historical data points on an algorithm that you create with Hero_Flow's building blocks. 

You begin with your known input data. You can create your own unique algorithm using Hero_Flow's preprocessors, layers, and postprocessors. The end point is your known output data.

After you have created your model, you can train it to generate predictions on new data coming based on previous data the model has already processed.

Open AI Studio

AI Studio can be opened by:

  • Creating a new AI model
  • Editing an existing AI model

Preprocessors

Preprocessors are Functions applied to the input data before being fed to the deep learning model.

Add a preprocessor

  1. To add a preprocessor in your algorithm, click the add (+) button between your input data and the first layer.
  2. A pop-up menu presents you with a choice a preprocessor.
    You can scroll and select your option or type in a search string to filter.
    Double-click your option or select it and click OK..
  3. The new preprocessor is added to your AI Studio algorithm.

Preprocessor list

PreprocessorDescription
Categorical
EmbeddingRepresent data as a vector.
Feature HashingCompute a uniformly distributed hash value for each category and assign each hash value to one of k buckets.
One Hot EncodingProduces a vector with length equal to the number of categories in the data set.
Numerical
BucketizingTurns numeric data into approximate categorical data.
Scaling

Rescale the data with the ability to shift it by an offset.

Other
Text EmbedderParts of pre-trained deep learning models that are used to transform text into rich vector representations.

Edit or delete a preprocessor

A preprocessor in AI Studio can be edited or deleted from the model process.

To edit a preprocessor:

  1. Right-click on the preprocessor box and select EDIT.

  2. View or update the configuration of the preprocessor.
  3. Click OK to save updates.

To delete a preprocessor:

  1. Right-click on the preprocessor box and select DELETE.

Layers

Layers are the mathematical processes used to build a deep learning model.

Add layers

  1. To add a layer in your algorithm, click an add (+) button between the dense layers.
  2. A pop-up menu presents you with a choice a layers.
    You can scroll and select your option or type in a search string to filter.
    Double-click your option or select it and click OK..
  3. If the layer has any configuration settings, a pop-up box is displayed.
  4. The new layer is added to your AI Studio algorithm.

Layer list

LayerDescription
Merge
Add

Layer that adds a list of inputs.

Average

Layer that averages a list of inputs.

Concatenate

Layer that concatenates a list of inputs.

Dot

Layer that computes a dot product between samples in two tensors.

Maximum

Layer that computes the maximum (element-wise) a list of inputs.

Subtract

Layer that subtracts two inputs.

Trainable
ConvolutionLayer that extract features from an input image.
DenseLayer that feeds all outputs from the previous layer.
EmbeddingLayer used on text data. Requires that the input data be integer encoded so that each word is represent by a unique integer. 
Long Short Term Memory LayerLayer used to process both single data points and sequences of data.
Recurrent Neutral LayerLayers used for connections between nodes to form a directed graph along a temporal sequence.
Utility
Average Pooling

Average pooling for temporal data.

Batch Normalization

Layer that provides a layer with inputs that are zero mean/unit variance.

BidirectionalLayers that duplicates the first recurrent layer so that there are two layers side-by-side. It then provides the input sequence as-is as input to the first layer and provides a reversed copy of the input sequence to the second.
DropoutLayer used to prevent a model from overfitting.
FlattenLayer that brings all levels of a multi-layered image down to one plane. 
Gaussian NoiseLayer that applies additive zero-centered Gaussian noise.
Language ModelLayer that provides probability distribution over sequences of words.
MaskingLayer that masks a sequence by using a mask value to skip timesteps.
 
Time DistributedApplies a layer to every temporal slice of an input.
UpsamplingLayer with no weights that doubles the dimensions of input and can be used in a generative model when followed by a traditional convolutional layer.

Edit or delete a layer

A layer in AI Studio can be edited or deleted from the model process.

To edit a preprocessor:

  1. Right-click on the preprocessor box and select EDIT.

  2. View or update the configuration of the layer.
  3. Click OK to save updates.

To delete a layer:

  1. Right-click on the layer box and select DELETE.

Postprocessors

Postprocessors are Functions applied to the output field rather than the input.

Add a postprocessor

  1. To add a postprocessor in your algorithm, click the add (+) button between your last layer and the output data.
  2. A pop-up menu presents you with a choice a postprocessor.
    You can scroll and select your option or type in a search string to filter.
    Double-click your option or select it and click OK..
  3. The new preprocessor is added to your AI Studio algorithm.

Postprocessor list

PostprocessorDescription
Categorical
One Hot EncodingRepresent data as a vector.
Numerical
Log TransformTransform the target data into its logarithm. 
ScalingRescale the data with the ability to shift it by an offset.

Edit or delete a postprocessor

A postprocessor in AI Studio can be edited or deleted from the model process.

To edit a postprocessor:

  1. Right-click on the postprocessor box and select EDIT.

  2. View or update the configuration of the postprocessor.
  3. Click OK to save updates.

To delete a postprocessor:

  1. Right-click on the postprocessor box and select DELETE.

AI Studio Toolbar

OptionDescription
SaveSaves the AI model in Hero Flow.
TrainBegins training the AI model. See Train an AI Model. 
ConfigureOpens the window to the AI model training configuration settings.
UndoUndoes the last command.
RedoRedoes the last command which was undone.
Reset ViewResets the view of the the AI model to the default view.
Training DashboardOpens a new window with information and comparisons for up to three model versions. 

Related Topics