Major NNTrack Update: New Recurrent Layers

New recurrent neural network layers are now available in NNTrack:

  • SimpleRNN
  • LSTM
  • GRU

The neural network designer can now be used to build hybrid CNN + RNN architectures that analyze images in greater depth.

Why does this matter?

After convolutional layers, a network produces a feature map of the image. Recurrent layers can process these features as a sequence, gradually analyzing the relationships between different parts of the image.

This helps a model:

  • take context between features into account more effectively;
  • analyze dependencies between image regions;
  • identify complex structures and details more effectively.

These architectures are especially useful for complex image classification tasks where it is important to consider not only individual features, but also their positions and relationships.

The new layers will be available in an upcoming version of NNTrack.

← All news