No compromises. Fully integrated. Totally dialed.
Having a powerful tool does not mean you should have to compromise on accessibility and ease of use. Synapse 2, the successor to Synapse combines an intuitive interface that allows you effortlessly create a complete solution from data visualization and model building to training and deployment. Every last bit of it does its job flawlessly and each piece fits together forming a complete integrated system. At your fingertips you have the world's most powerful algorithms running on your computer or in the cloud across one or thousands of CPU:s or GPU:s.
Understand, transform, define
Preprocessing in Synapse gives you a ridiculous amount of power to visualize and transform your data. Visualizers can be connected chains for unlimited drill down data mining. Non-destructive filtering allows you to modify and transform your data. Selections allow you to pick out interesting subsets of your data - either through a powerful query language or by visual selection (or both). Do you find yourself missing writing code? No problem - there's a filter for that! Enjoy the journaling function that creates automatic documentation for your filters, selections and graphs. It even documents automatically what you did to get a certain graph.
Breakthrough AI effortlesly
Build the most powerful systems in the world or just a bit of logistic regression. It's all the same to Synapse. You can just as easily create massive convolutional neural network and LSTMs or something as simple as a a bayesian classifier or a logistic regression model. Or why not try a combination? Synapse gives you an unprecedented ease of use by simply visually connecting blocks or just inserting snippets from a rich library. No need to worry about execution on GPU vs CPU or local vs Cloud - all that is taken care of automagically.
Train here, train there or train everywhere
Sometimes it makes sense to train a model on your laptop, but often you want to train it on a powerful set of GPUs or on a HPC machine. Synapse 2 has been designed from the ground up with the goal that it is not something you should have to worry about. You can run it on your computer, your workstation, your server or if you want we can run it for you in the cloud on powerful GPU powered servers. Just press play and you're off. Want to train a bunch of systems and optimize them? Just use the integrated optimizer (Genetic Algorithms, Particle Swarm optimization etc). You can monitor and control the training within Synapse or just 'fire and forget' with cloud logging to an ELK stack.
Analyze, Test, Verify then do it all over again
For any real world tasks, model tests and verification are a necessity. Synapse provides all the statistical and computational tests you wanted but was afraid to ask for. A large number of statistical tests and visualizations make sure that you can have confidence in your model. You can send the results of the tests back for further tuning of the data and the model.
One click production deployment. Seriously.
One click is literally all it takes to deploy a system to a Peltarion Spine 2 server (your own or in the cloud). A RESTful interface gives you direct access to your model. Apart from easy integration, Spine provides version control, automatic scheduled training and parallel execution of different models. Spine was developed in cooperation with several banks and has proven track record of extreme robustness in real world critical production systems. Best of all: we can run it in the cloud for you, so it doesn't have to be more complicated than a click.
Sign up for beta: