In this video:
- Romeo Kienzler, Chief Data Scientist, IBM
Romeo begins with some staggering figures about devices online, now and in the near future (up to 20 billion devices by 2020). IBM has invested heavily, in money and personnel, in IoT, creating the Watson IoT global headquarters in Munich. IBM has also contributed technology to the field, creating the open source protocol, MQTT, which has become the standard protocol for IoT.
Romeo next reviews some general facts about machine learning, leading up to a discussion of neural networks. Neural networks are built in layers which, over time, become more and more numerous, and pile up deeper and deeper, hence the term “deep learning.” As one goes deeper into the layers, they become less abstract and more concrete. Romeo illustrates this by describing how neural networks treat human faces. The input layers tend to be abstract, mostly just shapes and angles. The output layers, however, as one goes left to right, take on a resemblance to the actual human face.
The visual arts are just one area where neural networks can be trained to create. Romeo shows examples of text written by an “electronic Shakespeare” and he plays music composed by machines.
What challenges there are to neural networks and deep learning can be and are being met. For example, they tend to be computationally complex, but this problem can be addressed by advancements in hardware technology, such as IBM’s TrueNorth chip.
The time of the Internet of Things is now.
Follow Romeo as he tackles the most difficult challenges in data science.