Digital Commons Network^™

Universal Retrieval For Recommendation System By Graph First Relational Learning, Anonymous

Defensive Publications Series

Existing retrieval (aka candidate generation) system for recommendation is highly fragmented, with typically 100s of heuristic-based candidate generators. Because the majority of the output from each generator is duplicated with those from other generators, most of the computation used for training and serving these generators are wasted. Maintaining such a fragmented highly complex system is also a significant debt. Moreover, because these generators are trained independently with each relying on a specific type of narrowly-focused heuristic, the performance of such a retrieval system is highly suboptimal.

GFR aims to provide a unified modeling paradigm for retrieval. By unifying and integrating …

Mesh Compression, Anonymous

Defensive Publications Series

Embodiments described herein improve the lossless compression ratio of the Corto codec. An example mesh used for 3D graphics rendering may have many vertices, where each vertex coordinate is composed of X,Y integer pixel position coordinates, and Z is a floating point depth coordinate whose value ranges from 0 (close) to 1 (far). The depth value is coded as a “half float” 16 bit value.

Power Stage Driving Schemes For Multilevel Converter, Anonymous

Defensive Publications Series

The disclosure is directed to an improved multilevel DC-DC converter, which is commonly used by electronic devices with subsystems that have different voltage needs. The improved DC-DC converter is more efficient and smaller than conventional converters.

Anomaly Detection In Compressed Video Streams Using Machine Learning, Yuriy Aleksandrovich Romanenko

Defensive Publications Series

Video outlier or anomaly detection is currently performed using computer vision and image processing techniques such as filtering, motion estimation, etc. that require decompressed video streams. Applying machine learning techniques directly to a video stream is computationally expensive. This disclosure describes the use of a pre-trained machine learning model that can analyze compressed video bitstreams directly without decompression to identify likely anomalies in the video. Such direct analysis of compressed video streams can be performed at significantly lower computational cost while still yielding a strong anomaly detection signal. Sequences in which anomalies are detected can be decompressed and further analyzed …

Dual Data Center Controller (D2c2), Raouf Khan, Hafeeza Fathima

Defensive Publications Series

Disclosed herein is a system and method for an automatic data center switching between primary data center 202 and secondary data center 204 handled by Dual Data Centre Controller D2C2 208, 216 in case one of the data centers becomes inaccessible. The method involves the creation of a new file/block at both data centers and placing a “.completed” file in the respective files of HDFS data centers. The method further involves checking the presence of “.completed” file at respective HDFS data centers along with the deadline time and then initiating the file transfer based on the deadline time and the …

Method And System For Metadata Maintenance In A Data Lake, Harleen Kaur, Durga Kala

Defensive Publications Series

The present disclosure provides a method and a system for identifying stale data sets in the data lake for deprecation. The system comprises a metadata Hub, having a processor and a memory, and may act as a central repository. The metadata hub may also include a deprecator system or a deprecator, that uses an AI (artificial intelligence) model to deprecate unused datasets. The metadata hub may be communicatively connected to a large-scale data repository, also known as a data lake. The metadata hub may be configured to control the flow of ingest data from any relational database into the data …

Enhancing Model Security: Leveraging User-Generated Ids As Embedded Watermarks In Machine Learning Models, Alan Gatzke, Mitchell C Mosure, Andi Wilson, Ananth Racherla, Todd C Kuehnl, Tyler Shane Mccarthy

Defensive Publications Series

The potential theft or unauthorized use of machine learning models developed by a company can lead to significant financial losses and damage to the company's intellectual property. While existing methods of protecting machine learning models such as encryption or access controls can be circumvented by skilled attacker, techniques presented herein involve the integration of embedded watermarks into machine learning models. Such techniques involving the integration of embedded watermarks may not only uniquely identify a model but may also include a unique user identification/identity that can make it possible to track usage of the model and detect any unauthorized use of …

Method And System For Controlled Information Disclosure Based On User Context Using Principles Of Just-In-Time And Just-Enough-Access, Sachin Cherian Mathew, Ying Teo

Defensive Publications Series

The present disclosure discloses a system and method to controlled information disclosure based on user context using principles of Just-In-Time and Just-Enough-Access, during package transportation. The system includes a unique identifier generator module for receiving transportation information, associated with a package, and generates a machine-readable unique identifier. The generated unique identifier is stored in a database and placed on the package. Upon receiving the package at the shipper sites, the unique identifier on the package is scanned using a scanning device with a GPS module and the information is provided to a computing unit, for computing and retrieving information about …

Language Assistant Capability Exchange Protocol, Hui Wang, Jayachandran C, Amruth Ramachandran

Defensive Publications Series

This disclosure describes techniques to support language assistance services for communication between devices. Per techniques of this disclosure, language assistant capabilities are exchanged between participating devices during call setup. A call protocol such as session initiation protocol (SIP) is suitably modified to exchange language assistant parameters during call set up. The SIP header fields are extended to add a parameter to indicate language assistant capabilities by indicating a suitable language code. An INVITE request is transmitted from an originating device that includes a language identifier that indicates the language assistant capabilities of the originating device. A corresponding language identifier of …

Factorized Blank Thresholding For Improved Runtime Efficiency Of Neural Transducers, Anonymous

Defensive Publications Series

We show how factoring the RNN-T's output distribution can sig­nificantly reduce the computation cost and power consumption for on-device ASR inference with no loss in accuracy. With the rise in popularity of neural-transducer type models like the RNN-T for on-device ASR, optimizing RNN-T's runtime efficiency is of great interest. While previous work has primarily focused on the optimiza­tion of RNN-T's acoustic encoder and predictor, this paper focuses the attention on the joiner. We show that despite being only a small part of RNN-T, the joiner has a large impact on the overall model's runtime efficiency. We propose to factorize the …

The Quad Sensor System That Collectively Collects Light Projections Utilizing A General Sensor Positional Configuration Where Each Sensor System Captures A Specific Distributed Area Of The Scene, Punarjeewa Abeysekera

Defensive Publications Series

This paper describes a configuration which consists of a quad sensor system that collectively collects light projections utilizing a general sensor positional configuration where each sensor system captures a specific distributed area of the scene. The configuration consists of a concave mirror, 4 identical beam splitters and 4 sensor systems. The quad sensor system will collectively have the capability to capture the light rays coming from a scene. The 4 sensor systems will utilizes a general sensor positional configuration. Each sensor system in the configuration will capture a specific distributed area of the scene.

Method And System For Detecting Fraudulent Transactions Through Payment Card, Mohan Kumar Sabapathy, Vijayraj Shanmugaraj, Vaishale Mahadevan

Defensive Publications Series

The aspect of the present disclosure relates to a method (500) and a system (122) for detecting fraudulent transactions through payment card (102). The method (500) initiates with collecting and storing a set of data associated with card holding patterns of a card holder (106) using a smart chip (202) of the payment card (102). The set of data is transmitted from a smart chip (202) to a POS terminal (110) and forwarded to a payment processor (112). The payment processor (112) compares the set of data with historical data to identify a change in the card holding pattern when …

Establishing Phone Number Reputation Using A Contact Center Platform, Claus Regenbrecht, K Srinivas, Akiko Sexton

Defensive Publications Series

Spam telephone calls decrease customer confidence in unknown callers; consequently, businesses can experience low call pickup rates. Some challenges with accurate caller identification are that a call from a legitimate business can be marked as spam, and that the name of the business that appears on the display need not be the actual owner of the number. This disclosure describes techniques to determine the accuracy of the caller ID and additional description such as business name or spam status reflected on a callee’s mobile device. An outbound test call is initiated from a contact center platform to a test mobile …

A Method For Establishing Card Relationships And System Thereof, Hunny Huria, Kira Blundell

Defensive Publications Series

The present disclosure relates to a technique for establishing a relationship between a plurality of payment cards held by the cardholder and further utilizing the established relationship to suggest alternative payment options, in case the first card is declined. The technique involves sending a query to a relationship manager (RM) by the merchant if the initiated transaction is declined to inquire whether it stores any established relationship between the cardholder’s multiple payment cards, communicated to it by the issuer entities. In response, the RM searches its database and deploys the learning model to suggest the best replacement option out of …

Personalized Room Detection, Anonymous

Defensive Publications Series

[0001] Techniques described herein provide an efficient technique for performing location-based tasks, which may be used by a variety of devices, including AR/VR devices. According to particular embodiments, an AR/VR device may selectively use localization techniques (e.g., SLAM) and machine learning (ML) to determine the location of the user, which in turn may be used to trigger downstream applications, such as personal assistant or recommendation services. Localization is used to determine the user’s specific 3D pose within the user’s environment, which would be useful for applications that need to know the user’s particular 3D pose. For example, in order to …

Flip Sapphire Light-Emitting Diode (Led) Chip Package, Suying Jin

Defensive Publications Series

This publication describes an improved flip sapphire light-emitting diode (LED) chip package. The improved LED chip package overcomes delamination, glue bleeding, insufficient glue, and other adhesion issues that are commonly found in prior LED chip packages. In the improved LED chip package, instead of using a glue stamping method to attach a phosphor layer on a flip sapphire LED, a die attach film (DAF) is utilized.

Dynamic Stack Port Serdes Power Utilization For A Sustainable Data Stack, Amitabh Ranjan, Rakesh Maduvinakodi Rangaswamy, Girish Kumar Gupta

Defensive Publications Series

Multiple network switches may be stacked, one atop another, and then interconnected through each switch’s two stack ports (SPs). Within such a stack arrangement, techniques are presented herein that support dynamically reducing the power utilization of a switch’s SP Serializer/Deserializer (SerDes) blocks without interrupting any data traffic. Under aspects of the presented techniques, a switch may transition between different power saving modes, which may include a normal mode (encompassing powering down the SerDes blocks of both of the switch’s SPs), an optimized mode (encompassing reducing the speed of the SP SerDes blocks), and a smart mode (encompassing dynamically powering up …

Systems And Methods For Tasks Based On Personalized Room Detection, Anonymous

Defensive Publications Series

A technique for performing location-based tasks by selectively using a machine-learning-based location detection module and a localization module, depending on whether downstream applications only need the user’s general location or the user’s 3D pose in order to accomplish an intended location-based task.

Adaptive Packet Acknowledgement

Defensive Publications Series

High performance AI applications require very high bandwidth network communication along with predictable tail latency. Any congestion on the network directly affects the tail latencies and thereby affects the performance of the end applications. A principle of an example solution is for the receiver timer to be a function of the rate at which the transmitter sends data packets and adjust based on the number of packets received since the last acknowledgement was transmitted. With this, the receiver timer can shorten and send frequent ACKs if very few packets are received to urge the transmitter to keep its tracker window …

Techniques To Facilitate Graphics Processing Unit (Gpu) -Aware Vector Packet Processing (Vpp) Flow Offloads For Edge Compute Resources, Dennis Lanov

Defensive Publications Series

Presented herein are techniques that address issues related to enabling a Vector Packet Processing (VPP) core for Artificial Intelligence (AI), Virtual Reality (VR), and/or Augmented Reality (AR) workload(s) on edge compute resources in which the edge compute resources can provide multiple edge applications. Some applications provided at the edge compute resources may utilize generic Central Processing Units (CPUs) while others may utilize Graphics Processing Units (GPUs).

A Process For The Preparation Of 1-(3-Chloro-5-{[4-(4-Chlorothiophen-2-Yl)-5-(4-Cyclohexylpiperazin-1-Yl)-1,3-Thiazol-2-Yl]Carbamoyl}Pyridin-2-Yl) Piperidine-4-Carboxylic Acid Or Its Pharmaceutically Acceptable Salts, Msn Laboratories Private Limited, R&D Center; Srinivasan Thirumalai Rajan; Sajja Eswaraiah; Sagyam Rajeshwar Reddy; Sheri Venkata Reddy And Batchanaboina Venkata Subbaiah

Defensive Publications Series

The present invention relates to a process for the preparation of 1-(3-Chloro-5-{[4-(4-chlorothiophen-2-yl)-5-(4-cyclohexylpiperazin-1-yl)-1,3-thiazol-2-yl] carbamoyl} pyridin-2-yl) piperidine-4-carboxylic acid which is represented by the following structural formula-1 or its pharmaceutically acceptable salts.

Formula-1.

Acid Addition Salts Of Belumosudil And Their Process For The Preparation Thereof, Msn Laboratories Private Limited, R&D Center; Srinivasan Thirumalai Rajan, Sajja Eswaraiah, Vijayavitthal T. Mathad,Saladi Venkata Narasayya, Kammari Balraju And Garai Abhijit

Defensive Publications Series

The present application relates acid addition salts of 2-{3-[4-(1H-indazol-5-ylamino)-2-quinazolinyl]phenoxy}-N-(propan-2-yl) acetamide of formula-1 and their process for preparation thereof.

Formula-1.

Cross Customers Smart Network Inventory Planner (Snip) And Optimizations Using Deep Reinforcement Learning, Pengfei Sun, Qihong Shao, Gurvinder Singh

Defensive Publications Series

Optimal inventory upgrade planning is one of the most challenging tasks in managing network assets. A Smart Network Inventory Planning (SNIP) architecture or framework is presented herein that leverages a deep reinforcement learning (DRL) framework to enable network inventory upgrade planning for different scenarios. As a foundation for the DRL framework, techniques herein provide for establishing a network inventory environment through which interaction with a supply chain can be used to allow the SNIP architecture to incrementally optimize upgrading sequences for multiple customers. To further optimize inventory upgrades via the DRL framework, the SNIP architecture may employ a multi-objective reward …

Adaptive Layer Height, Rijk Van Manen, Oscar Van De Ven

Defensive Publications Series

This publication relates to a method of slicing a 3D object wherein a layer height is adapted according to the local geometry of the object. The proposed solution will enable a 3D printer to print an object with an improved dimensional accuracy on the horizontal surfaces.

Safe-Send, Sachin Cherian Mathew Visa, Chang Kang Charles Goh Visa, Irvin Sie Ze Ting Visa, Jyoti Sahu Visa, Wen Yuan Galvin Leow Visa, Jing Rong Chan Visa, Loh Guo Yang Visa

Defensive Publications Series

The present disclosure relates to a method of performing a Safe-Send transaction that may allow a sender to recall money transferred to a receiver even after the transaction is completed. According to the method, when the sender sends a transaction request the sender’s bank configures a Safe-Send limit and a Safe-Send time limit for the sender to perform a recall of the transaction. This is based on a concept called “Future-Lien.” So, once the money is transferred to the receiver’s bank account, the money is under Lien and the receiver would not be able to use the money for the …

Blue Green Deployment Strategy For Applications Using Kafka, Rinkle Seth, Pragya Priyadarshini, Rushil Sharma, Vrinda Rastogi, Yash Dwivedi, Asheesh Gupta

Defensive Publications Series

BLUE GREEN DEPLOYMENT STRATEGY FOR APPLICATIONS USING KAFKA

Disclosed herein is a blue-green deployment strategy for consumer applications using the Kafka system. The present disclosure proposes setting producer and consumer configurations in a way that green application consumers are able to consume messages from Kafka topic, hence enabling application teams to test their changes in a green environment. To achieve this, the present disclosure provides a rebalancing of partitions technique where the number of partitions are configured either twice the number of consumers or greater than twice the number of consumers so that all consumers in a consumer group are …

Payment Reminder Technique With Enhanced Efficiency, Alok Roy Visa

Defensive Publications Series

The present disclosure relates to the payment reminder process. It proposes to improve the efficiency of payment reminder process by sending the reminders at customer’s preferred timing (like at what time in the day, on which day of the week, on which date of the month) and communication channel (like email, SMS, phone call). A reinforcement learning technique and particularly Q-learning may be used to derive the decisions based on parameters like customer’s payment history, past reminder interactions at different timings, reminder channels (like email, SMS, phone call); and find the best possible timing and channel of communication for the …

Visa Buy Later Pay Later, Varun Sharma, Ajit Vilasrao Patil

Defensive Publications Series

The present disclosure relates to an automated, secure and reliable payment platform that allows recurrent transactions. Particularly, the present disclosure relates to system and method for supporting recurring payments for retail and commercial purchases in a four party arrangement. The method comprises: initiating (S310) a Visa Schedule Program (VSP) for transactions based on the user’s schedule(s); providing (S320) the user 110 an ability to create, program, and cancel the schedule(s); and generating (S330) an intuitive program schedule 280, due to which the card is not stored as card on file at merchant location, preventing fraud.

Dynamic Location Determination Through Short Range Neighbor Discovery, Craig Wickham, Dave Zacks, Cat Nguyen, Leila Reyhani, Nandini Maheshwari

Defensive Publications Series

Techniques are presented herein that support a close-proximity, ultrasound-based discovery and data exchange capability that provides a non-intrusive method for exchanging data (such as, for example, location information) between devices that share the same airspace. Using ultrasonic sound, information may be shared between any device capable of transmitting and/or receiving ultrasonic sound waves and elements of a network or building infrastructure allowing for the exchange of very specific and dynamic location data. One immediate use case for such technology is in the handling of emergency service (e.g., 911) calls where time and accuracy are critical. The presented techniques may find …

System And Method For Managing Faults In A Distributed Environment, Vrinda Rastogi, Seth Rinkle, Rushil Sharma, Pragya Priyadarshini, Yash Dwivedi, Asheesh Gupta

Defensive Publications Series

The present disclosure discloses a method and a system for managing faults in a distributed environment 102. In the present disclosure, the method includes monitoring health metrics of systems 106 in the distributed environment 102. Further, the method includes detecting faults associated with the systems 106 in the distributed environment 102 by identifying abnormal patterns based on monitored data. Furthermore, the method includes reconfiguring the distributed environment 102 to maintain system resilience and performance based on fault detection. Further, the method includes determining a recovery action based on severity of faults. Furthermore, the method includes analyzing and diagnosing issues by …