Neuropype 2023: Our Biggest Release Yet
It's time for our next major release (2023) to our NeuroPype signal processing suite! This is our biggest release to date, with major new features and literally hundreds of new nodes that continue to expand NeuroPype's capabilities and use cases.
A major new component is the ability to create and run deep neural networks inside NeuroPype pipelines, with convolution layers, dense layers, normalizations, activations, optimizers, schedules, cross-validation, training, and inference. Comprehensive deep learning models, such as EEGNET can now be expressed as a NeuroPype pipeline. You can now create end-to-end pipelines from data curation, preprocessing and feature extraction, to deep learning predictions and cross-validation. These pipelines can also process multiple modalities (i.e., EEG, Gaze, other biosignal data) and incorporate them into the same model.
NeuroPype now supports GPU processing, with full support for JAX, Tensorflow and Pytorch. Most ML/DL, and math operation nodes, support multi-GPU processing out of the box, but will also work on the CPU for non-GPU machines.
NeuroPype also now supports non-linear workflows such as looping, conditional processing, and allows you to define re-useable functions. Many new nodes have been added to represent programming constructs such that you are now able to create Turing complete pipelines! This greatly expands the ability to do processing in a graph pipeline without the need to do some programming in python to "glue" parts together. You can still create your own custom nodes, but can do a lot more with the nodes already there.
In all we've added over 300 new nodes, bringing the total to over 800(!). This includes some completely new node packages (categories): Deep Learning, Distributions, Optimizations, Random, Programming, Array, and Control Flow, plus many new array-handling nodes in the ElementWise package, new nodes for working with independent components (in the Neural package), new statistics nodes such as MANCOVA, multiple comparison corrections, etc., in the Statistics package, and many others. Many nodes were also enhanced, and bugs fixed too.
Of particular note, and dear to our heart, NeuroPype is now available for Linux! (You'll find the Linux downloads in the User Guide, under Downloads, which you can access from your NeuroPype account.)
Pipeline Designer has also seen some upgrades to make it more user-friendly and helpful particularly in terms of exposing documentation to better understand your pipelines, nodes and their properties, with support for HD screens, easier graph manipulation (zoom, pan, search), etc. Enjoy!
Check out the Release Notes for details, and grab the latest release today by logging into your Neuropype account, or if you're not a Neuropype user, sign up today. We greatly appreciate any feedback you may have, especially those of you using the Academic Edition.
