Microsoft Technology Licensing, LLC

Context Relevant, Inc.

Salesforce Com, Inc.

Salesforce, Inc.

James E Walsh

Andrew Kadatch

Sameer Tiwari

Dustin Lundring Rigg Hillard

Stephen Purpura

Ethan Nelson Ray

Hugh Evan Williams

Stuart Raymond Patrick

Kenneth Alan Moss

Arthur Zwiegincew

Julia H Farago

Ariel J Lazier

Nicholas A Whyte

Kavi J Goel

Philip Fung

Brian D Wentz

Jonathan A Forbes

Ronald A Fein

Michael Ronn Marcelais

Xiaowen Shan

Ebbe H Altberg

Benjamin Aaron Rudiak-Gould

J Scott Stamps

Jonathan M Dubman

In a computing environment, a set of executing processes each having associated resources are provided. Aggregate resources for the computing environment include multiple different types of resources. A utilization level for each of the resources within the computing environment is evaluated to determine an unconsumed capacity for each of the resources below a utilization threshold. The utilization threshold is resource-dependent. An indication of at least a portion of unconsumed capacity for each of the resources below the utilization threshold is gathered. The unconsumed portion for each of the resources below the utilization threshold is exposed for consumption by other executing processes.

Techniques and architectures for efficient allocation of under-utilized resources

A computing environment having a first plurality of executing processes being executed by one or more processors. The executing processes each have associated resources. The executing processes utilize the associated resources. A consumed portion of a resource that is utilized by a selected executing process from the first plurality of executing resources is determined. An unconsumed portion of the resource is estimated. A first preselected amount of the unconsumed portion is allocated for utilization by a process not in the first plurality. A second preselected amount of the unconsumed portion is reserved as a fence between the consumed portion and the allocated first preselected amount of the unconsumed portion.

Techniques and architectures for protection of efficiently allocated under-utilized resources

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a machine learning model. One of the methods includes receiving a complete set of training data; receiving instructions to train a predictive model having a plurality of parameters on an initial subset of the complete set of training data; training the predictive model on the initial subset; storing data representing a first state of the predictive model after training the predictive model on the initial subset; receiving updated parameter values and instructions to train the predictive model on a new subset of the complete set of training data; and training the predictive model on the new subset.

Training a machine learning model

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a machine learning model. One of the methods includes performing experiments to select a training strategy for use in training the model on a particular data set. The selected training strategy includes a binning strategy for binning the raw feature vectors before the raw feature vectors are provided to the predictive model.

Selecting a training strategy for training a machine learning model

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training machine learning models. The models can include models for predicting a next transaction price or a next transaction price direction for one or more financial products, for classifying particular debit or credit card transactions as likely being anomalous or fraudulent or not, or for classifying particular financial claims processing transactions, e.g., insurance, health care, or employee expense claims transactions, as likely being anomalous or fraudulent or not.

Re-training a machine learning model

A system and method are provided for detecting artificially generated load on a search system. The system may include a load monitoring component for monitoring a current load for comparison with an expected load. The system may additionally include an abnormality detection component for detecting an abnormality when the monitored load exceeds an expected amount by a predetermined threshold. The system may further include an analysis component for determining if the monitored load is an artificial load.

Detection of artificially generated system load

A unique object navigation system, user interface, and method that facilitate faster and smoother navigation of objects are provided. For example, the system can generate a plurality of objects that can be rendered on a display space that spans a single page in length, thereby mitigating the need to navigate through multiple pages. The system can determine the length of the page according to the number of objects generated. To view off-screen objects, an infinite scroll component can be employed. The amount of scroll space needed to view the plurality of objects can be determined in part by the length of the page and/or by the number of objects. The objects can also be viewed in a film strip format that is infinitely scrollable. The film strip view allows a view of the objects to be maintained while also viewing a selected object in greater detail at the same time.

Object search UI and dragging object results

System and method for detection of artificially generated system load

Object search ui and dragging object results

Adaptable accelerated content streaming allows data to be streamed from a server to a client over a network at an accelerated rate for an amount of time before reducing the rate. In accordance with one aspect, the accelerated rate is prohibited from exceeding a threshold amount, which is based at least in part on both a total bandwidth of the server and an amount of bandwidth currently being used by the server. In accordance with one aspect, the data can be streamed to the client at up to an adaptable accelerated rate.

Adaptable accelerated content streaming

An improved hibernation method and system, including the use of a modified DMA (Direct Memory Access) mode of transferring data to and from the disk. The use of DMA increases data transfer speed, while freeing the system processor to perform other tasks, including compressing/decompressing the data transferred to and from the disk. An improved decoder is also provided that reduces the number of bounds checks needed on average for typical compressed data by first guaranteeing that there is sufficient room to decode literals and small substrings, whereby bounds checking is not needed. A combination hibernation mode and a suspend mode is also provided that essentially maintains power to the RAM while transparently backing the RAM with the hibernation file, such that if power to the RAM is interrupted, the RAM contents are automatically restored from the hibernation file when power is restored.

Method, system, and program for hibernation of computer systems having DMA controller to asynchronously read data from buffers

A system and method are presented for monitoring user browsing information. Such information can include, but is not limited to, the web pages visited by users, search queries submitted by users, the manner in which users browse the Internet and search for content, as well as any demographic information and interests of the corresponding users. Once a particular type of user browsing information has reached a certain threshold of activity by users, the invention can be configured to detect activity that reaches the threshold and then can increase the monitoring of the information.

Real-time analysis of web browsing behavior

The present invention leverages high-frequency interrupts and/or low priority threads to accurately determine which computing resources are available. This provides a computing asset (CPUs and/or software applications) with a means to accurately compensate for resource utilization in order to increase its performance. By utilizing the present invention, the computing asset can optimize its performance in a real-time, self-tuning manner. In one instance of the present invention, high intensity, low priority threads are initiated on available CPUs (logical and/or physical) to effectively replace a CPU's idle time with the low priority thread. This thread generally constitutes a computationally-intensive and/or a memory-intensive thread which permits a highly accurate performance measurement to be obtained for available CPU resources. In another instance of the present invention, high-frequency interrupts are initiated on CPUs to compensate for processes that are shorter than a thread's time quantum, providing a more accurate performance counter.

High performance counter for realistic measurement of computer system load

A method is provided for storing and providing responses to keyword search queries and other queries. One of a plurality of memory devices for storing a query answer is selected based on a hash value of the query. The hash value is used to select a permutation for ordering the memory devices. One or more memory devices for storing the query answer are selected based on the ordering. Additionally, stored query answers are updated based on comparing a storage time to one or more threshold values.

Growing community of inventors

Woodinville, WA, United States of America

James E Walsh