|Title||Machine Learning Hardware Case Study for Smart Buildings|
|Publication Type||Technical Report|
|Year of Publication||2020|
|Authors||Chai, E, Murmann, B|
|Place Published||Stanford University|
|Keywords||Design, Edge hardware, Machine learning, smart hospitals|
The potential impact of deep learning on the healthcare industry is extraordinarily high. The size of this impact is raised by the size of the healthcare industry, one of the largest industries in the United States by GDP and expected to grow from 17.9% in 2018 to 19.7% by 2028 . The CDC projects healthcare costs to rise due to the anticipated increase in the size of the aging population . The result will be an increased strain on the current public healthcare systems, especially if the number of healthcare workers cannot meet this growing demand. This strain will exacerbate the prevalence of unintended patient harm from human error, which was already a well-known problem in the industry .
Deep learning, combined with new possibilities in smart buildings and work environments, can mitigate this strain via applications in the smart hospital environment and has already inspired research in this direction. For example, recent work at Stanford  (Albert Haque, Serena Yeung, Arnold Milstein, Fei-Fei Li et al.) looked at vision-based smart hospitals and demonstrated a system for monitoring the staff’s hand hygiene. While the achieved results are impressive, taking a system from a proof-of-concept demonstrator to a scalable smart building solution requires considerable progress in hardware/software codesign. It is naïve to expect that deep learning algorithms can be easily moved from the academic environment, with curated data sets such as ImageNet , to the hospital environment at scale, without any complications or roadblocks. This work highlights the smart hospital environment’s constraints and guides future AI researchers to apply AI in a practical way to this environment.
|Alternate Title||ML Hardware Case Study for Smart Buildings|