Recent years have witnessed an explosive growth of AI models. The high cost of hosting AI services on GPUs and their demanding service requirements, make it timely and challenging to lower service costs and guarantee service quality. While spot instances have long been offered with a large discount, spot preemptions have discouraged users from using them to host model replicas when serving AI models.

As distributed applications become increasingly complex, so do their scheduling requirements. This development calls for cluster schedulers that are not only general, but also evolvable. Unfortunately, most existing cluster schedulers are not evolvable: when confronted with new requirements, they need major rewrites to support these requirements. Examples include gang-scheduling support in Kubernetes [6, 39] or task-affinity in Spark [39]. Some cluster schedulers [14, 30] expose physical resources to applications to address this. While these approaches are evolvable, they push the burden of implementing scheduling mechanisms in addition to the policies entirely to the application.

Hyperparameter tuning is essential to achieving state-of-the-art accuracy in machine learning (ML), but requires substantial compute resources to perform. Existing systems primarily focus on effectively allocating resources for a hyperparameter tuning job under fixed resource constraints. We show that the available parallelism in such jobs changes dynamically over the course of execution and, therefore, presents an opportunity to leverage the elasticity of the cloud.

Prior research in resource scheduling for machine learning training workloads has largely focused on minimizing job completion times. Commonly, these model training workloads collectively search over a large number of parameter values that control the learning process in a hyperparameter search. It is preferable to identify and maximally provision the best-performing hyperparameter configuration (trial) to achieve the highest accuracy result as soon as possible.

Emerging smart cities are typically equipped with thousands of outdoor cameras. However, these cameras are usually not calibrated, i.e., information such as their precise mounting height and orientation is not available. Calibrating these cameras allows measurement of real-world distances from the video, thereby enabling a wide range of novel applications such as identifying speeding vehicles and city road planning . Unfortunately, robust camera calibration is a manual process today and is not scalable. In this article, we propose AutoCalib, a system for scalable, automatic calibration of traffic cameras. AutoCalib exploits deep learning to extract selected key-point features from car images in the video and uses a novel filtering and aggregation algorithm to automatically produce a robust estimate of the camera calibration parameters from just hundreds of samples. We have implemented AutoCalib as a service on Azure that takes in a video segment and computes the camera calibration parameters. Using video from real-world traffic cameras, we show that AutoCalib is able to estimate real-world distances with an error of less than 12%.

Videos have ample amount of information in the form of frames that can be utilized for feature extraction and matching. However, face images in not all of the frames are “memorable” and useful. Therefore, utilizing all the frames available in a video for recognition does not necessarily improve the performance but significantly increases the computation time. In this research, we present a memorability based frame selection algorithm that enables automatic selection of memorable frames for facial feature extraction and matching. A deep learning algorithm is then proposed that utilizes a stack of denoising autoencoders and deep Boltzmann machines to perform face recognition using the most memorable frames. The proposed algorithm, termed as MDLFace, is evaluated on two publicly available video face databases, Youtube Faces and Point and Shoot Challenge. The results show that the proposed algorithm achieves state-of-the-art performance at low false accept rates.