The task of traversing the near limitless combination of input features, classifier types, and associated parameters is one well-suited to automation with results that typically outperform the heuristics and insight of a seasoned expert performing this task by hand. SensiML’s segmentation, feature pre-processing, and transform libraries includes over 80 routines built for efficient execution in the IoT sensing edge node.
Depending on selected target hardware, SensiML Analytics Studio leverages specific platform resources to improve efficiency with examples including use of Arm CMSIS DSP and Quicklogic FPGA acceleration. Resulting models (known as SensiML Knowledge Packs), can be assessed in bit exact device emulation of the model within Analytic Studio and delivered in binary, object, or full source code form.*
* Full source code output requires Standard or Enterprise level subscription.