Thursday, April 25, 2019 - 1:30pm to 2:45pm
Location:7501 Gates Hillman Centers
Speaker:PRAHALADHA MALLELA, Masters Student /PRAHALADHA%20MALLELA
A Novel, Dynamic, Adaptive Machine Learning Layer for IoT
Internet of Things environments are widespread and are equipped with a variety of sensor data which can be used for machine learning applications. However, in IoT settings, the environment itself is not static but changes over time, leading to variations in the sensor data. Furthermore, in complex environments, applications are exposed to various new conditions over time. Each IoT environment also has unique sensors and devices present. This makes machine learning applications in the IoT environment challenging.
This thesis presents an IoT centric, end-to-end Machine Learning Layer which addresses these challenges. The ML Layer architecture enables each of the aspects (training and prediction serving) to feed into one other leading to a continuous cycle. It includes a flexible model definition that allows it to incorporate any type of model or framework. Initially, it optimizes models, and performs ensemble model selection. Over time, as a part of its autonomous feedback loop it is able to automatically identify different patterns in environmental data, and continuously adapt models based on this feedback. In addition, the system optimizes dimensionality reduction based on environmental data in the long run to improve efficiency. The system is designed to be general to accommodate any type or combination of data.
The Machine Learning Layer for IoT is also a fully managed service, designed to be flexible and adaptive so that it can create a powerful plug and play interface. It can be deployed in a variety of settings (including smart homes and smart cars) which require specialized learning on the spot to fit the environment and continuously improve accuracy.
Yuvraj Agarwal (Chair)