Cogito Instruments provides a full suite of sub-systems, software tools and services to let you embed intelligence in your systems.
Cogito Instruments first hardware subsystem product is the CI9120. It is a CompactRIO® cartridge for National Instruments® platform. It contains 1024 neurons, giving your system the ability to learn up to 1024 patterns and then recognize and classify them. The CI9120 is able to perform pattern recognition and classification in 10µs. Incoming patterns from the neural network are delivered by the CompactRIO® controller through the CompactRIO® backplane connector. Outgoing classification responses from the CI9120 neural network are sent back to the controller through the backplane connector as well. The content of the neurons (knowledge) can be downloaded and uploaded through the backplane connector.
THANKS TO THE CI9120, YOUR CompactRIO® BASED MACHINE WILL BE ABLE TO:
Generate a decision space, thanks to the CI9120 adaptive model generator
Perform multidimensional non-Linear data analytics, thanks to the CI9120 high performance classifier using Radial Basis Function (RBF) or K-Nearest Neighbor (KNN)
Proliferate knowledge. At any moment, the internal knowledge of the neurons can be saved and stored for later use or for proliferation to other systems.
Cogito Instruments provides the software tools, based on LabVIEW®, enabling an easy integration in your applications. Thanks to a simple API using LabVIEW® virtual instruments, you can:
- Acquire your signals within your CompactRIO® system (from a camera, a sensor, or data transfer from a host)
- Extract the recognition features and format your data into 256 Bytes long vectors
- Send those vectors to the CI9120
- First for teaching, using reference vectors (constructed from reference signals). Each time you teach a vector, you can associate a category to it. This tells the neurons what they should learn. Each time a new vector is learned, the neurons automatically refine their knowledge.
- Then for recognition. The module will answer whether the vector is recognized or not. If it is recognized, it will return the associated category (class) of the matching vector. In some cases, the result can be flagged as “uncertain” if more than one category is matched. In this case you can check how many neurons were involved in the recognition of each category, and even get a “distance” measurement for each neuron to decide which one has the most meaningful result, that helps refining the matching criteria with learning on the fly.
- At any moment, the internal knowledge of the neurons can be saved and stored for later use. Once the teaching phase is over and the system properly recognizes all submitted vectors, the internal knowledge can be exported for replication onto other systems.
Cogito Instruments provides technical support and H/W and S/W customization services to help you integrate this ground-breaking technology in your systems. We also have a network of partners who are able to provide assistance and system integration services.