Digital Commons Network^™

Active Directory And Calendaring Based Allocation And Reservation Of Wireless Resources, Sukrit Dasgupta, Jp Vasseur, Santosh Pandey

Defensive Publications Series

Present techniques dedicate wireless channel resources based on applications that will be used in collaborative meetings. The degree of reservation is determined using machine learning algorithms that have knowledge from a prior history of meetings, participants in the meeting, etc. This ensures that collaborative applications will perform favorably, leading to more effective collaborations. Several novel components are involved in these techniques, which control the network infrastructure, determine the users that need reservation(s), identify the applications that will be used, and also utilize machine learning to model the usage and requirements.

Device Orientation Detection Through Signal Features, Xu Zhang, Amir Kamalizad, Khashayar Mirfakhraei

Defensive Publications Series

Techniques are described herein for using Radio Frequency (RF) signal features at multiple Access Points (APs) to detect the orientation of a client device. These orientation-related RF signal features may be used over time and over multiple APs to perform device tracking/authentication and pattern recognition in 5G or Internet of Things (IoT) networks.

Human Usage Description For 5g Network Endpoints, M. David Hanes, Chuck Byers, Joe Clarke, Gonzalo Salgueiro

Defensive Publications Series

Just as Manufacturers Usage Description (MUD) defines security and access control for “things,” Human Usage Description (HUD) does the same for humans by associating a user and all of their devices in a database. With such a database, a 5G provider has insights into user behavior on the network in context with the user device. This allows for more accurate Quality of Experience (QoE) and network slicing at the user device level, common licensing and security across multiple devices, next generation media consumption, and revenue generation through targeted advertising.

Automatic Failure Triage And Recovery Using Unsupervised Learning, Jon-Paul Sullivan

Defensive Publications Series

Techniques are provided to improve pipeline automation and reduce mundane operational tasks by automating the triage and remediation of any pipeline failures. This increases the efficiency of operations engineers and allows them to spend their time performing more useful activities. Unsupervised learning means that over time the effectiveness of the solution improves, and the nature of the solution enables it to scale better than a human would at the same task.

Matching A Graph With A Non-Deterministic Finite Automaton, Alexandre Boulgakov

Defensive Publications Series

The problem of matching a graph with a non-deterministic finite automaton (NFA) is of importance in various domains of computer science. An example is regular-expression matching, which can be formulated as a graph-matching problem. Current techniques of matching graphs against NFAs have relatively high computational complexity. This disclosure presents matching techniques with complexity that is linear in the size of the graph. The graph to be matched against the NFA is itself considered as an NFA. A synchronized product of the two NFAs is defined, and the matching problem is shown equivalent to a reachability problem solvable in linear (time …

Refined Music Clustering, Christian Weitenberner, Ullas Gargi, Girum Ibssa, Brian Karlak, Krishmin Rai

Defensive Publications Series

Many different versions of the same song are often found in online music or video services. To enable easier search and better user experience, online music services often use clustering techniques to cluster different versions of the same song. Clustering systems are designed to be tolerant to shifts, noise, small tempo changes, etc. This tends to cause remixes and covers of a song to be clustered with the original versions of that song. This leads to certain problems, e.g., recommender systems selecting the most popular version of a song, in preference to other versions of the song; poor precision-recall rates …

Soc Against Electric Vehicles, Daniel Hoppe

Defensive Publications Series

SOC AGAINST ELECTRIC VEHICLES

Lambdaloss: Metric-Driven Loss For Learning-To Rank, Cheng Li, Michael Bendersky, Marc Najork, Xuanhui Wang, Nadav Golbandi

Defensive Publications Series

How to directly optimize ranking metrics such as Normalized Dis­counted Cumulative Gain (NDCG) is an interesting but challenging problem, because ranking metrics are either flat or discontinuous ev­erywhere. Among existing approaches, LambdaRank is a novel algo­rithm that incorporates metrics into its learning procedure. Though empirically effective, it still lacks theoretical justification. For ex­ample, what is the underlying loss that LambdaRank optimizes for? Due to this, it is unclear whether LambdaRank will always converge. In this paper, we present a well-defined loss for LambdaRank in a probabilistic framework and show that LambdaRank is a special configuration in our framework. This framework, …

Progressive Delivery Of Interactive Content, Jared M. Zimmerman

Defensive Publications Series

Interactive elements in online content often lead to substantially increased total payload size to be delivered to a client device, e.g., to render a page. Delivering such content may introduce latency, thus degrading user experience. The techniques of this disclosure address this issue by delivering a compressed static image that mimics the visual feel of interactive elements of content. The image can be delivered faster because its size is significantly smaller than that of the interactive content. Subsequently, the image is replaced with the corresponding interactive content. With user consent, the replacement of the image with the corresponding content may …

Traffic Redirection With Distributed Denial Of Service Segment Identifiers, Robert Barton, Jerome Henry, Siva Sivabalan

Defensive Publications Series

Distributed Denial of Service (DDoS) information is extended to a network Path Computation Element (PCE). The PCE uses a network function Segment Identifier (SID), referred to herein as a DDoS SID, imposed on the edge routers by the PCE, to identify potentially suspicious DDoS traffic. The DDoS SID is used in Segment Routing (SR) routers to direct suspicious traffic to nearby or specially optimized DDoS scrubbing engines so that traffic may be cleaned. Other traffic flows proceed unchanged through the network.

Dynamic Wireless Local Area Network Access Control List Provisioning, Chethan Channappa, Mahesh Satyanarayana

Defensive Publications Series

Techniques are described herein for optimizing Ternary Content-Addressable Memory (TCAM) space through a digital network architecture controller by preventing Access Control List (ACL) rules from being downloaded to TCAM until a client is associated with that Service Set Identifier (SSID). The space on the TCAM is freed once the last client on that particular SSID has disassociated.

Digital Campaign Tracking With Respect To Application Infrastructure, Piyush Mathur, Bhupinder Singh, Siddhartha Dutta, Arpita Agarwal

Defensive Publications Series

Digital campaigns are often derailed from their goals due to amiss or often miscalculated aspects of the impact of the campaign on application infrastructure. To mitigate this problem, an approach is described herein which unifies these otherwise disjoint domains of digital marketing and application monitoring analytics. A solution is provided whereby a user may perform capacity planning for upcoming campaigns and perform real time correlation on campaign and infrastructure Key Performance Indicators (KPIs), thus maximizing revenue and optimizing infrastructure cost. Through machine learning algorithms which are running on top of correlated data from both the domains, a plethora of actionable …

Side-Effect Free Program State Evaluation, Erik Luo, Aleksei Koziatinskii, Yang Guo

Defensive Publications Series

The ability to inspect the state of program execution at a specified stage by the evaluation of a specific expression is crucial for operations such as debugging. Executing the expression of interest may itself result in changing the program state that the evaluation of the expression is supposed to inspect. Such side effects of evaluating the expression on the program state reduce the effectiveness and reliability of debugging by evaluating the expression to inspect program state. The techniques of this disclosure enable evaluation of an expression for inspecting program state that is free from side effects.

Woptimop - A Cloud Based Intelligent Method Of Automatically Creating And Delivering Workload Optimized Platform Contents, Suprateeka R Hegde, Sujoy Saraswati, Amrita H S

Defensive Publications Series

UNIX/Linux being the most preferred enterprise operating platform, provides state-of-art compiler and developer tool chain for almost all programming languages. And so far, has greatly helped enterprise application developers to optimize their applications for their specific workloads, through various set of options and tunables.

However, the solution is still partial, as only the application is optimized for the workload. The Operating System and other components of the platform, though allows tuning and customization, is common for all. Hence each and every application and workload being characteristically different, cannot benefit much from such operating platforms.

Here we propose a unique innovative …

Machine Learning Prediction System Based On Tensor-Flow Deep Neural Network And Its Application To Advertising In Mobile Gaming, Leon Li

Defensive Publications Series

This disclosure analyzes the background of the industry and proposes solutions to the challenges faced by complex applications or games, such as strategy games. From the perspective of the system architecture, this disclosure describes how to clean up data, identify salient features, model predictive classifiers, and automate analysis and selection for content delivery. Data collection and processing have a great influence on the accuracy and applicability of the model. Four kinds of behavioral parameters (or more) may be used to predict conversion events, with PCA used to reduce the dimensions utilized as inputs to the model. In addition, by adjusting …

Cutting Tool For Trench Stabilization Channels, John Korman, D. Clay Cox, David Walsh

Defensive Publications Series

This disclosure describes a cutting tool for cutting horizontal stabilizer channels along sidewalls of trenches. The cutting tool includes a shank with a depth stop, a bearing guide, and a cutting head. The cutting head is configured to make a horizontal cut into the sidewalls of the trench at a specified depth. The depth can be adjusted to accommodate different engineering requirements. The cutting tool is driven by a vertical shaft from an electric or gasoline powered walk behind router. The cutting tool can be utilized to provide horizontal stabilizer channels in trenches that can reduce repair costs due to …

Real-Time Scheduling For Software Testing, Keun Soo Yim, Yuexi Ma

Defensive Publications Series

Software is subjected to a large variety of tests on different devices prior to public release, including pre-submit and post-submit tests, that have different scheduling attributes. The number of physical devices available for testing is limited, e.g., due to the devices being unreleased or under development. This disclosure provides techniques for scheduling software tests on a scarce pool of physical devices. The tests have differing scheduling attributes, e.g., durations, periodicity, degree of preemptiveness, etc. Tests are scheduled such that a test experiences low latency (queueing delay) regardless of test duration, the availability of the device pool is improved, and device …

Secure Audio Processing, Thad Hughes, Ignacio Lopez Moreno, Aleksandar Kracun, Pedro Moreno

Defensive Publications Series

Automatic speech recognizers (ASR) are now nearly ubiquitous, finding application in smart assistants, smartphones, smart speakers, and other devices. An attack on an ASR that triggers such a device into carrying out false instructions can lead to severe consequences. Typically, speech recognition is performed using machine learning models, e.g., neural networks, whose intermediate outputs are not always fully concealed. Exposing such intermediate outputs makes the crafting of malicious input audio easier. This disclosure describes techniques that thwart attacks on speech recognition systems by moving model inference processing to a secure computing enclave. The memory of the secure enclave and signals …

System For Optimally Shipping Items To Minimize Costs, Jeffrey Cuartero

Defensive Publications Series

A system and method are disclosed that optimally ships items to minimize costs. The method includes receiving by the system an order for the items to be packaged. The system determines various parameters for the order, to provide a parcel recommendation to a packer, which may include determining a possible carrier and rate structure, dimensions, number of items, location, weight of the items, etc. The system uses a suitable optimization model to minimize cost of shipping. The optimization output may be order dimensions including size and weight, and carrier for each package. The system then instructs the packer to pack …

Safe Parking: Further Development Epb +, Daniel Hoppe

Defensive Publications Series

Safe Parking: Further Development EPB +

Adaptive Audio Processing Based On Scene Detection, Leonardo Kusumo, Moonseok Kim

Defensive Publications Series

This disclosure describes techniques to automatically adjust audio processing parameters based on acoustic scene detection. A parameter adjustment module utilizes default dynamic range compression settings during initial audio recording using a mobile device. A trained machine learning model is utilized to detect the acoustic scene, based on initial audio signals. Features such as the average loudness, maximum loudness, minimum loudness, acoustic scene, wind noise, and other user context features, as permitted by the user, are provided to the trained model. The model is utilized to adapt audio processing parameters such as the compression profile to the detected scene to improve …

A Methodology For Automated Feature Engineering In Training Of Shallow Neural Networks For Prediction Of Multi-Linear Growth Failures (“Nova Algorithm”), Dermot Cochran, Martin Roe, Cv Goudar, Breda Martyn, Mainak Das, Pratush Uppuluri, Sean Lennon, Basudev Panda, Sarah King, Liam Dolan, James Marshall Hughes

Defensive Publications Series

Deep learning requires a large data set. When working with smaller data sets we need an automated approach for feature learning as a pre-processing step to create the training set for a shallow neural network with one hidden layer. The outcome of the neural net classifier is filtered using post-processing rules over a defined wait and quiet period to create an ensemble prediction with higher quality. This gives us the benefits of deep learning, but with a smaller data set.

A partial implementation of this methodology was used for a pilot release for prediction of Common Previsioning Group (CPG) growth …

Automation And Orchestration Of Hardware And Firmware Data Mining Using A Smart Data Analytics Platform, Suhas Shivanna

Defensive Publications Series

Effective data mining is going to be important for differentiating and succeeding in the digital economy especially with increased commoditization and reduced barrier to entry for infrastructure devices like servers, storage and networking systems. There is lot of telemetry data from manufacturing facilities and customers that can be used to drive improved supportability experience, unmatched product quality and reliability of infrastructure devices like servers and storage devices. Currently data mining of hardware, firmware and platform logs is a challenging task as the domain knowledge is complex with expertise for large multinational organization distributed across the world. With increasing complexity and …

Polite Mode For A Virtual Assistant, Thomas Deselaers, Pedro Gonnet

Defensive Publications Series

The use of imperative statements and commands for a voice-activated virtual assistant can set an inappropriate example to children and can potentially lead them to imitate such language in normal conversation. However, current voice-activated virtual assistants are not configured to recognize polite language which can lead to unintended responses. This disclosure describes a virtual assistant that can be configured in a polite mode. The disclosed techniques may be utilized to configure a virtual assistant such that it is responsive only to polite queries. Polite mode is enabled by use of voice stress analysis, sentiment analysis module and natural language understanding …

Net Neutrality Value Pack Using Network Data Analytics, Vidya V. Bharadwaj, Prashanth G.A, M Mahesh Babu, Suhasan Chirravuri

Defensive Publications Series

The advent of mobile internet and the phenomenal growth of the use of smart phones has brought data onto the forefront, creating newer revenue streams for the operators. The data/Internet connection now needs to cater to diverse traffic, just as a city must manage the flow of various vehicles and pedestrians on its streets. In the data world, usage of data ranges across various applications like streaming-video, real time gaming, B2B & M2M applications. Such diverse customers often blame their operators for throttling data flows to the phones or computers. This causes significant delays and losses in data transmission. Any …

Spreadsheets With Programmatically Accessible Version History, Tal Cohen

Defensive Publications Series

Spreadsheets enable users to store and track values of a variable. To graph changes in values of a variable over time, a user typically stores in the spreadsheet values of the variable at different time points. Retaining all values in individual, visible cells of a spreadsheet can make the spreadsheet difficult to read and/or maintain.

Modern spreadsheet programs include features to track changes made to the cells of a spreadsheet, e.g., as a version history. The version history enables a user to revert to a prior version. Techniques of this disclosure enable programmatic access to historical values stored in the …

Device For Achieving A Redundancy In, Daniel Hoppe

Defensive Publications Series

Device for achieving a redundancy in

the wheel speed sensor failure esp. For

piloted driving in traffic (fail operational)

Brake Recuperation / Increase In Braking, Daniel Hoppe

Defensive Publications Series

BRAKE RECUPERATION / INCREASE IN BRAKING

POWER DUE TO THERMAL STORAGE

Incremental Sharing Using Machine Learning, Victor Cărbune, Sandro Feuz, Daniel Keysers, Thomas Deselaers

Defensive Publications Series

Generally, the present disclosure is directed to incrementally sharing resources, e.g., documents, media, etc., using machine-learned models. In particular, in some implementations, when the user provides permission to access user data, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to suggest files on a device or in the cloud that a user can share based on documents or other media that the user has already shown intent to share.

Machine learning is used to determine resources, e.g., documents, media, links, etc., that have features with correlation with the resource the …

Identifying Outages In User-Facing Products Based On Search Statistics, David Coles, Yehuda Karlinsky, Gil Ran, Shlomo Urbach

Defensive Publications Series

A method is disclosed for identifying outages in user-facing products based on search statistics. The method includes collecting search queries related to a user-facing product from a search engine. The method further includes generating statistics based on the collected search queries. The statistics may indicate spikes based on the product and failure-related terms associated therewith. Based on the statistics, the spikes for searches related to outage of the product are identified. Further, on identifying the spikes, alerts may be issued to an appropriate technical team related to the outage.