The number of wireless sensors connected to the IoT is large, and growing exponentially. These devices will help monitor a broad range of conditions, from soil moisture to people’s health. Of course, the presence of these sensors is only half the story – the data they generate will grow exponentially too, and will be a key component of deep learning, allowing machines to tackle challenges beyond anything we’ve seen thus far.
Analyzing New Content Types
In years past, when it came to machine learning, text was king. Processing text files helped form the basis for algorithms. While that will continue, the scope of learning is about to get a lot broader. Karen Panetta, IEEE Fellow and Dean of Graduate Engineering Education for the School of Engineering at Tufts University, predicts “Deep learning will help us analyze and better utilize the massive amounts of data being shared. We are not just analyzing text anymore. Image content and audio analysis will become instrumental to making the IoT impactful.”
Analyzing audio and video opens up learning opportunities that can better approximate the human experience. For Shawn Chandler, IEEE Senior member and director at Navigant, potential applications include “various types of artificial neural networks (ANN), particle swarm optimization (PSO), genetic and evolutionary computing, agent based simulation, dynamic programming and many others.”
As these new approaches are vetted and verified with real data, they’ll improve the function of a number of industries, and subsequently our quality of life.
Far from Simple
The idea that massive amounts of data coming from the IoT will translate into gains in deep learning is not a guarantee. In fact, as the scale of data from connected devices rapidly multiplies, finding ideal data sets to teach machines could become a serious challenge. To Sukanya Mandal, IEEE member and Data Science Professional in India, some advanced planning is necessary: “We should be concerned about making the maximum utilization of IoT data for generating smart insights. Solutions needs to be built so that the most data can be fed into the analytics pipeline, while less of it is lost due to redundancy.”
Changing Other Aspects of Life
Deep learning stands to complement our behaviors and inform our decision making, especially on complex problems. Panetta sees our electricity use as an obvious starting point: “Monitoring our power consumption will be the next big feature, helping to create profiles of our needs and where we spend our time. Not just how much we use (as shown on our electric bill), but where and what devices we spent most of our time using.”
IEEE recently convened 19 experts on artificial intelligence, machine learning and cybersecurity, and they collectively authored a paper on the challenges society is facing. You can read more about it here; Transmitter will have a series of articles on it in the near future.
Written by IEEE on June 6, 2018