This paper 1 studies how to schedule hyperparameters improve generalization of both centralized single-machine stochastic gradient descent (SGD) and distributed asynchronous SGD (ASGD). augmented with momentum variants (e.g., heavy ball (SHB) Nesterov’s accelerated (NAG)) has been the default optimizer for many tasks, in environments. However, advanced variants, despite empirical advantage over...

Existing transport protocols for periodic information dissemination ignore application semantics while attempting to be 100% reliable. \Application Level Framing" (ALF) suggests that taking application semantics into account when designing transport protocols can result in application performance that is highly optimized for the network. We apply this principle in designing a policy-based TUNAb...

This paper presents work on piggyback cache validation (PCV), which addresses the problem of maintaining cache coherency for proxy caches. The novel aspect of our approach is to capitalize on requests sent from the proxy cache to the server to improve coherency. In the simplest case, whenever a proxy cache has a reason to communicate with a server it piggybacks a list of cached, but potentially...

Caching at proxy servers is one of the ways to reduce the response time perceived by Web users Cache replacement algorithms play a central role in the response time reduction by selecting a subset of documents for caching so that a given performance metric is maximized At the same time the cache must take extra steps to guarantee some form of consistency of the cached documents Cache consistenc...

Traditionally, decision tree classifiers work with data whose values are known and precise. We extend such classifiers to handle data with uncertain information. Value uncertainty arises in many applications during the data collection process. Example sources of uncertainty include measurement/quantization errors, data staleness, and multiple repeated measurements. With uncertainty, the value o...

Modern society is heavily dependent on wireless networks for providing voice and data communications. Wireless data broadcast has recently emerged as an attractive way to disseminate data to a large number of clients. In data broadcast systems, the server proactively transmits the information on a downlink channel; the clients access the data by listening to the channel. Wireless data broadcast...

Highly distributed data management platforms (e.g., PNUTS, Dynamo, Cassandra, and BigTable) are rapidly becoming the favorite choice for hosting modern web applications in the cloud. Among other features, these platforms rely on data partitioning, replication and relaxed consistency to achieve high levels of performance and scalability. However, these design choices often exhibit a trade-off be...

Business Intelligence (BI) is a set of techniques that help improve business decision making. From a technical point of view, BI relies on a set of tools which includes performance dashboards: layered services that combine monitoring, analysis, and reporting. However, most dashboard solutions today are based on data warehouses which face a problem of data staleness—a circumstance caused by the ...

We present SMART, a scalable, bandwidth-aware monitoring system that maximizes result precision of continuous aggregate queries over distributed data streams. While previous approaches reduce bandwidth cost under fixed precision constraints, in practice, monitoring systems may still incur a substantial cost risking overload under bursty traffic conditions. SMART therefore bounds the worst-case ...