With the intentions of ending the week well, I decided to go through Mozer’s paper. Our seniors and professor considered this to be a model paper given the extensive use of reinforcement learning by the researchers.
I read this paper written my a few researchers in University of Michigan. The name Newman drew me in, since I have seen the name multiple times in this field. This paper was like many others: a study on how people interact with a smart home system they came up with.
I was quickly skimming through the paper since I was quite well-aware of what most HCI papers talk about, but slowed down a bit when I found an interesting approach to testing how well their system OSCAR works. OSCAR is a form of end-user programming that allows user to flexibly control their devices of networked devices at home. The paper did not mention this explicitly, but from a photo, I could tell that OSCAR in a physical form is a tablet that people can interact with.
After a quick but brain-boggling look at Machine Learning yesterday, we returned to more HCI oriented papers today. I chose to read CAMP: A Magnetic Poetry Interface for End-User Programming of Capture Applications for the Home (Truong et al.).
I read a paper written by Microsoft on HomeOS. Just from reading, I do have to give credit to them for being pioneers in the field. We do know that there are some difficulties using their system and devices, but their innovative idea on PC abstraction is indeed very promising for smart homes and user satisfaction.
Microsoft advocates centralized PC operating system for home and decentralized network of devices because 1)all devices will be connected and this connection will allow for smart home functionality, 2)it is easy to add devices since they are not closed system, 3)users already know how how to use computers, and 4) it is easier for developers to implement functionality without worrying about devices.
Yesterday we’d asked Professor Littman for a text on machine learning. The result of our query was a hefty tome – Artificial Intelligence, a Modern Approach by Peter Norvig (one of my role-models) and Stuart Russell. The concepts were straightforward enough to understand but the text was littered with forehead-wrinkle-inducing jargon (eg: classification, regression, realizable hypothesis spaces). I managed to get through a section or two, stopping at ‘learning decision trees’; it was slow going especially since I was making a personal (metaphor filled) summary on my research wiki.
I read a paper called “Toward User-Centric Feature Composition for the Internet of Things“ written by Zave, Cheung, and Yarosh. This was a good paper to start with because it defined what “Internet of Things” stand for and what is expected out of it.
Internet of Things in a broad sense stand for a smart home where sensors and actuators are connected through the Internet. So far, there are more complaints than satisfaction because people expect the system to behave “intuitively” which may not be so natural for computers. On top of that, each individual is a complex being distinct from others, which makes it hard to come up with a model that fits everyone. Continue reading
Today I waded through a paper outlining the shortcomings of the Nest smart thermostat and its limited success at learning users’ preferences. I also read a long rant from a tech blogger who had tried and failed at basic home automation. The combined (albeit limited) picture I got from the two articles conjured a somewhat bleak status quo for machine learning and user-programming in smart homes. Which of course, is just what an eager young researcher wants to hear (I’ll also try reading some more upbeat literature tomorrow).