Machine Learning for IoT applications

Machine Learning for IoT applications

Why would Artificial Intelligence (AI) be used in IoT Applications?

The adoption of AI within embedded systems is gaining momentum in many different areas from airplanes and drones to robots, industrial sensors and biomedical monitors. Focusing for instance on biomedical monitors, what is the benefit of Embedded AI? When embedding Predictive Models within monitors measuring vital information, the system can process right away incoming data and immediately issue predictions, such as potential health problem. No time waisted transferring vital data records back and forth to the cloud, no risk of data security breach and highly limited latency.

All these steps are carried out locally, in the monitor itself, almost instantaneously.

Problems to solve

  • What can Machine Learning bring to Embedded Systems?
  • How can an Embedded System self generate a prediction based on its own onboard data?
  • For instance, is it possible to add to a cardiac monitor a stroke prediction feature?
  • Is it simple and easy to embed such a prediction feature into an embedded system in general?
  • Benefits of TADA
    for embedded systems

    Embedded Systems experts and Healthcare Professionals can benefit from AI without being a data scientist. They usually do not have the required skills in machine learning nor in software coding to build predictive models. Most data collected by these professionals are in the range of a few hundreds to a few thousands. The data collected for each patient can be alarms from electrocardiogram (ECG) signal, heart and breathing rates, temperature, walking speed… These Data typically fall within the category of Small Data, meaning a few hundreds or thousands of patients data but rarely millions (aka Big Data). Traditional machine learning algorithms work well with Big Data but do not perform well with Small Data.

    MyDataModels allows Healthcare Professionals to automatically build predictive models from their Small Data. Raw data can be used directly without any preprocessing: no normalization, nor handling of outliers or feature engineering are required. The predictive results from processing these specific datasets can be obtained in a few clicks in less than a minute on a standard laptop.

    Thanks to its light footprint of 2KB once trained and modelled, a TADA software prediction engine can be embedded into the memory of most microcontrollers.

    Therefore, to process a predictive model, a device does not need to send the data out into the cloud to be computed. It can do it locally in the end equipment.

    By processing the data close to the sensors that produce it, network latency times are avoided, providing important gains in responsiveness and safety, as opposed to a more traditional cloud computing approach.

    Data security concerns are avoided thanks to this local approach. In applications such as predictive maintenance and healthcare, manufacturers may be unwilling to transfer their key data in the cloud for analysis.

    MyDataModels brings a self-service solution for those who have Small Data and no Data Scientists.

    TADA brings new possibilities for IoT on embedded systems

    Now, with TADA, the Embedded and Healthcare world can combine the use of IoT and machine learning to detect health problems more accurately and faster than ever.

    This symbiotic use makes it possible to train the model offline in the cloud or in a datacenter and then use it in the embedded system.

    The combined uses of IoT, will make machine learning faster than ever

    The main benefit of this mode of operation is that you can take advantage of advanced machine learning techniques to build models while being able to execute them locally on an embedded target. Taking a decision, monitoring or raising an alert locally on an embedded device within seconds may avoid serious health complications or even save lives.