Large-scale industrial recommender systems are usually confronted with computational problems due to the enormous corpus size. To retrieve and recommend the most relevant items to users under response time limits, resorting to an efficient index structure is an effective and practical solution. The previous work Tree-based Deep Model (TDM) [34] greatly improves recommendation accuracy using tree index. By indexing items in a tree hierarchy and training a user-node preference prediction model satisfying a max-heap like property in the tree, TDM provides logarithmic computational complexity w.r.t. the corpus size, enabling the use of arbitrary advanced models in candidate retrieval and recommendation. In tree-based recommendation methods, the quality of both the tree index and the user-node preference prediction model determines the recommendation accuracy for the most part. We argue that the learning of tree index and preference model has interdependence. Our purpose, in this paper, is to develop a method to jointly learn the index structure and user preference prediction model. In our proposed joint optimization framework, the learning of index and user preference prediction model are carried out under a unified performance measure. Besides, we come up with a novel hierarchical user preference representation utilizing the tree index hierarchy. Experimental evaluations with two large-scale real-world datasets show that the proposed method improves recommendation accuracy significantly.

No abstract

No abstract

Quantifying building energy performance through the development and use of key performance indicators (KPIs) is an essential step in achieving energy saving goals in both new and existing buildings. Current methods used to evaluate improvements, however, are not well represented at the system-level (e.g., lighting, plug-loads, HVAC, service water heating). Instead, they are typically only either measured at the whole building level (e.g., energy use intensity) or at the equipment level (e.g., chiller efficiency coefficient of performance (COP)) with limited insights for benchmarking and diagnosing deviations in performance of aggregated equipment that delivers a specific service to a building (e.g., space heating, lighting). The increasing installation of sensors and meters in buildings makes the evaluation of building performance at the system level more feasible through improved data collection. Leveraging this opportunity, this study introduces a set of system-level KPIs, which cover four major end-use systems in buildings: lighting, MELs (Miscellaneous Electric Loads, aka plug loads), HVAC (heating, ventilation, and airconditioning), and SWH (service water heating), and their eleven subsystems. The system KPIs are formulated in a new context to represent various types of performance, including energy use, peak demand, load shape, occupant thermal comfort and visual comfort, ventilation, and water use. This paper also presents a database of system KPIs using the EnergyPlus simulation results of 16 USDOE prototype commercial building models across four vintages and five climate zones. These system KPIs, although originally developed for office buildings, can be applied to other building types with some adjustment or extension. Potential applications of system KPIs for system performance benchmarking and diagnostics, code compliance, and measurement and verification are discussed.

No abstract