Tata Consultancy Services Patente

Embodiments herein provide a method and system for current density maps reconstruction by jointly optimizing x-direction, y-direction and z-direction magnetic field maps. Herein, the method of reconstruction of current image is leveraging state of the art iterative reconstruction techniques along with deep learning techniques. The quality of the current reconstruction depends on the signal to noise ratio (SNR) of the magnetic field observed. Since the noise model is difficult to characterize, a deep neural network is trained to learn the complex noise model from the data itself. To simulate the realistic scenarios, training data is created by introducing various kinds of noise that may appear in the system and collecting the training data pairs, i.e., pair of noisy and denoised x-direction and y- direction current density maps. The DNN is trained on this paired data to learn the noise model.

The embodiments of the present disclosure herein address unresolved problems of unavailability of a scalable framework to enable large spectrum of optical response-design combination with reduced computing time. Further, there are limitations in the choice of features to obtain the desired sensitivity. Embodiments herein provide a method and system for a generative artificial intelligence (GenAl) based sensor design synthesis of metamaterial. Herein, the sensor design synthesis framework is based on deep generative models. The GenAl based design synthesis model would provide that unknown information that would immensely be useful for optimizing the design towards the highest sensitivity, addressing the limitations in the choice of features using numerical simulator and easiest fabrication-feasibility following sensitivity response (SenR). The SenR is the sensitivity relationship between the physical properties of the ambient medium and the property of the sensor.

Monitoring and detection of oedema severity is crucial for effective treatment of the underlying ailments. Conventional methods involve setups that require harmful radiation, are expensive, tedious, and bulky. The present disclosure provides a system and method for unobtrusive oedema severity classification and quantification using microwave sensing. A vector network analyzer, a microstrip RF patch antenna customized for operation at 4 GHz under dielectric loading of a numerical human phantom and associated RF components are used for oedema severity monitoring of a sample under test. For oedema severity classification, a machine learning (ML) based approach is implemented making use of reflection parameters, and accordingly physics-based electromagnetic features are extracted. The physics-based electromagnetic features are used to classify the sample under test into a specific oedema severity category. Further, a water percentage is quantified using regression models for the sample under test based on the oedema severity category.

Polarization is a parameter that plays a pivotal role in communication during transmission and reception of signals. Presence of lumped elements along a signal path or around a region of interaction of incoming waves with surface may trigger distortion in the signals, leading to loss or damage of information. Generally, electronic lumped components are used but their functional abilities are limited during practical operations. The present disclosure addresses the unresolved problems of the conventional system by providing a system and method for real time multi-frequency operation using metasurface-based mechanically controlled tunable polarization conversion. In the present disclosure, a multi-band polarization converter design is provided whose operational parameters including (i) bands of operation, (ii) bandwidth, and (iii) number of bands are controlled together by modulating an air-gap of a reflective unit cell of a two-dimensional array structure of a metasurface using a motorized mechanical arrangement.

Existing therapeutic methods lack interactivity and adaptability to individual needs. The present disclosure retrieves from a database, a hardware component design as an interactive table comprising a grid of sensor(s) to detect input(s) from a user. The detected one or more inputs from user are passed to a gaming engine connected to the hardware component. Reflex time of user is recorded, when user provides the inputs to the sensors. Reflex time is compared with predefined time T_n and at least one of following is performed based on comparison that includes incrementally changing position of object, resetting timer, increasing timer with a first delta value and moving position of object closer to initial position with a second delta value. An average response time, a progression metric and adjustments to timer are calculated. A cognitive function to assess the severity of cognitive decline of user is derived.

A method and system that builds a regression model from a personalized structural and functional Cardiac Digital Twin (CDT) for assessment of cardiopulmonary endurance of an athlete is disclosed. The personalized Cardiac Digital Twin (CDT), which replicates echo like functionality under dynamic conditions integrates subject specific kinematics data real time acquired to run personalized CDT and generate intrinsic metrices to evaluate performance in different phases of exercise or endurance activity. Most of existing works are focused on computing mere metrices for entire activity as whole. However, without judicial combination of these metrices obtained in different phases, no meaningful inference can be drawn on performance evaluation.

Unrestricted access to large language models (LLM) based services can lead to potential data leakages, especially for large enterprises providing products and services to clients that require legal confidentiality guarantees. However, a blanket restriction on such services is not ideal as these LLMs boost employee productivity. Objective of the present disclosure is to build a solution that enables enterprise employees to query such external LLMs, without leaking confidential internal and client information. QueryShield platform of the present disclosure is a platform that enterprises can use to interact with external LLMs without leaking data through queries. It detects if a query leaks data and rephrases it to minimize data leakage while limiting the impact to its semantics. A language model is chosen from a set of lightweight model candidates that are identified and fine tuned for this purpose using a huge dataset and evaluated using multiple metrics.

Existing techniques of constructing classical equivalents of QML models expand measurements from QML model as truncated Fourier series and find Fourier coefficients by solving an optimization problem. However, this is computationally expensive and since approximate Fourier coefficients are determined, the resulting classical surrogate does not necessarily give exactly same predictions as that of the QML model. Embodiments of the present disclosure provide a method and system for constructing classical equivalent of trained Quantum Machine Learning (QML) model by distilling knowledge from the QML model to a classical model. The method includes training QML model using a first dataset, creating a second dataset using measurements from the QML model, training classical model using the second dataset and learning a processing function using predictions of the classical model and target outputs. The trained classical model combined with the learned processing function serve as classical equivalent of the QML model.

This disclosure relates generally to processing of gas using packed bed column, and more particularly, a packed bed column with stacked packed channels for processing gas. In current packed bed column configurations, the problems of maldistribution of gaseous stream across the packed beds and channeling associated with the packed beds results in high power requirements and operating expenses. The present disclosure includes gas distribution mechanisms and stacked packed channels for proper gas distribution and channeling of the incoming gaseous stream. The gas distribution mechanisms ensure that the incoming gas gets distributed evenly to the packed bed. This greatly increases the productivity of the process and increases the process efficiency as more area is available for heat and mass transfer to take place. The disclosed configuration is used for gas separation via adsorption, thermal energy storage applications and so on.

This disclosure relates generally to a method and system for local interpolation function (LIF) based synthetic aperture radar (SAR) imaging for irregular scanning geometries. Current available methods are computationally expensive, sensitive to initialization and lacks generalization, particularly when applied across sampling schemes in continuous sampling scale scenarios which is a key characteristic of irregular scanning geometries. The present disclosure achieves enhanced SAR image across various spatial coordinates. Noisy SAR images are obtained using Non-uniform Fast Fourier Transform (NUFFT) based Range Migration Algorithm (RMA). The noisy SAR image is mapped into a LIF feature map using an encoder of a trained LIF neural network. An enhanced SAR image is further predicted based on the LIF feature map using a decoder of the LIF neural network. The disclosed method for SAR image enhancement using LIF neural network is used for supply chain management, airport security screening, medical imaging and so on.

This disclosure relates generally to method and system to generate spatio-temporal intercrop layouts using 3D-CNN-LSTM based large vision model. The method is a pretrained spatio-temporal crop interaction model based on a 3D-CNN-LSTM to recommend optimal intercrop combinations. The method initially receives a user request as input which is preprocessed by removing noise and performing normalization. Further, a pretrained 3D-CNN-LSTM model is utilized to generate an optimal spatio-temporal intercrop layout for each intercropping scenario. The method also provides providing dynamic visual insight for the spatio-temporal intercrop layout comprising crop growth and corresponding intercrop interaction. Finally, an user feedback is obtained from at least one of the user or non-invasively in response to the user request for iterative improvements and fine-tuning the 3D-CNN-LSTM model using reinforcement learning.

This disclosure relates generally to method and system for identifying frauds in unemployment insurance claims using hybrid weighted decision model. The method combines predictive power of machine learning models with domain knowledge infused key fraud indicators and network analysis to identify false positive claims. Initially, a set of claim information from a request of a claimant is extracted to assess risk affecting eligibility of the claimant to receive benefits. Further, a classification probability for a set of key fraud indicators are predicted for each claim using a set of top features associated with a prescient artificial intelligence (AI) model. Finally, one or more frauds associated with the unemployment insurance claim of the claimant are identified based on the weighted probability of each claim, a network diagram generated using the weighted probability for each claim, a set of filtered claims, and a set of business rules.

The disclosure generally relates to methods and systems for multi-frequency fusion for localization of impact induced soundwaves of a target industrial machine. Conventional wideband beamforming-based techniques accurately localize the impact hotspots, however with increased computational time while beamforming-based techniques reduce the computation time but worsens the localization accuracy. In the present disclosure, a composite acoustic map and a single-tone acoustic map of each of the one or more single frequencies obtained at each predefined frequency interval, are generated based on geometric parameters and a characteristic sound signal. Then, the one or more localization regions associated with each experimental trail that are common to the composite acoustic map and the single-tone acoustic map are determined. Lastly an optimal frequency range is determined using a localization loss function, a Kalman filtering technique, and a probability distribution function based on the localization regions, for localization of impact hotspots.

This disclosure relates generally to method and system to perform interval analysis in source code using functional approach. The method performs data-flow analysis using functional approach to solve infinite-height analyses and properties are validated such as array index within bounds, non-zero division, and preventing arithmetic overflow or underflow on real-life applications. The method receives source code comprising one or more functions to perform a whole program interval analysis over each function using a functional approach to identify range interval of each variable at every program point in the source code. Further, summary for each function is computed which is stored in the form of variable and its corresponding range interval at every exit point of each function. Finally, an incremental interval analysis is performed over each function having edited version change and summary is recomputed for each function impacted by change to optimize overall interval analysis.

Modified Atmosphere Packaging (MAP) extends food shelf-life but is not customized for specific commodities and supply-chain conditions, leading to sub-optimal usage. The present disclosure utilizes physics-based models to simulate the system of MAP and generate data for shelf-life of the food commodity for various designs of MAP films stored in different environmental storage conditions. Further, this data is used to develop a hybrid recurrent neural network (RNN) and deep artificial neural network (ANN) model, which learns how the changes in storage conditions of temperature and relative humidity affect the shelf-life of the food commodity for different designs of MAP. Whereas optimization of multiple ordinary differential equations is time and compute resource intensive, the developed hybrid RNN-ANN model can be used as a surrogate which makes optimization significantly less resource intensive. A surrogateassisted optimization model thus developed can be used to predict the optimum design of MAP for specific storage conditions.

Conventionally, validating a configured jacquard card is manual, difficult and time-consuming process. Automated techniques require complex and costly hardware setup. Embodiments of present disclosure provide a method and system to verify and reconfigure design values of reconfigurable jacquard cards. It obtains a design from user and generates reference images of patterns required to weave the design. The user then configures the jacquard card based on the reference images and scans them after configuration. Patterns in the scanned images are validated against the reference images by identifying position of first top left hole and iteratively verifying the remaining holes by traversing one hole at a time and comparing its pixel value with that of the reference image. If the user has to create a new design using the same set of configured jacquard cards, order of the cards to be used for reconfiguring the new design with minimal cost is suggested.

This disclosure relates generally to processing of gas using packed bed column, and more particularly, a split bed packed column for processing gas. In current packed bed column configurations high pressure drop in packed column leads to increased energy consumption and puts a restriction on using high feed velocities, resulting in high cycle time. This results in increased requirement of adsorbent for processing of gas which thereby increases the operational expenses. The disclosed configuration includes multiple beds inside the same column differentiated by partitions instead of a single long bed inside the column. This configuration enables decreasing the pressure drop along the length, resulting in decreased energy consumption for the same cycle time, without compromising on other process key performance indicators. The disclosed configuration is used for gas separation via adsorption, thermal energy storage applications and so on.

This disclosure relates generally to a method and system for optimal power allocation in large-scale electric vehicle (EV) charging network having cluster topology. Currently, methods based on mathematical and constraint programming provide solutions for optimal power allocation but are computationally slow on large problem sizes. Reinforcement learning (RL) based learning-to-optimize (L2O) are also used for solving, but the solutions these methods provide impact the solution quality. The present disclosure provides mathematical programming-based method for optimizing power allocation decisions. The power allocation problem for the EV charging network is modelled with constraints on energy demand, power supply capacity, available network capacity, customer types and their priority level. Further the optimization problem is formulated to maximize the power flow to prime customers in the EV charging network. The problem is formulated using a mixed-integer linear program.

Conventional payment systems require payers to select the payment instrument first and then authorize payments. The payer cannot change payment instrument post authorizing payment as authorization is specific to a value store. The system and method of the present disclosure enables a payment method to be payment instrument/value store agnostic, so that payer gets more flexibility in terms of selecting which value store(s) to use for authorization but not necessarily for honoring the payment. This gives the payer the required flexibility to swap value store(s) post initial authorization but before settlement basis better informed decision by means of recommendations. Similarly on the payee's side, the payee has limited or no visibility of value store until payer presents it for payment. The system and method of the present disclosure gives an added opportunity to the payee to influence payer's selection of value store(s) post initial authorization but before settlement.

Log data recorded during digital assessments has hardly been ingested for analyzing log information to derive behavioral patterns that enable malpractice detection. A method and system for candidate response analysis during timed interval of digital assessment using statistical approach. The method provides a two-stage approach wherein the first stage analyzes raw log data of candidates to extract a set of attributes capturing activities of candidates in response to the questions appearing in the digital assessment. Unique statistical analysis is performed on the extracted attributes to determine outlier candidates that reflect anomalous behavior. At the second stage the set of outlier candidates are analyzed and filtered based on a plurality of checks against events that contribute towards possible malpractice or unfair means involved before and during exam duration. Thus, the outlier candidates that map to any one of the checks are identified to be attempting exam with Unfair means (UFM).

Brain monitoring is performed using an Ear-electroencephalography (Ear-EEG) which records synchronized electrical behaviour of neurons in brain at high temporal resolution using a wearable ear device. For proper functioning of the wearable ear device, sensor electrodes need to be within the ear to remain in constant contact with the skin. However, body movements tend to disrupt extent of contact, leading to noisy signals being captured, which are often difficult to distinguish from a valid EEG signal. Thus, identification of channels that are capturing EEG is necessary. The present disclosure provides a method and system for channel identification using manifolds of an Ear-EEG signal. In the present disclosure, a plurality of manifold features are extracted from a signal received from the wearable ear device. The plurality of manifold features are further classified as EEG and non-EEG channels using an unsupervised clustering model across real-world stimuli in the wearable ear device.

Development of a new drug molecule on average requires an expenditure of US$2-3 billion and a duration of at least 13-15 years. Subsequently, repurposing the existing drugs outside the scope of the original medical indications, has attracted serious consideration. Present disclosure provides a system and method for recommending drug candidates for a target disease by generating drug-refined Medical Subject Headings (MeSH) terms frequency matrix using MeSH terms assigned to Pubmed articles. Further, topic modelling is performed on the drug-refined MeSH terms frequency matrix to generate drug-topic conditional probabilities, and drug-drug distance is computed using Kullback-Leibler divergence. The drugs are then clustered and unknown drugs of a target disease are shortlisted from a cluster with known drugs based on predicted targets and indications of unknown and known drugs. A subset of shortlisted unknown drugs is recommended as a new candidate for target disease based on supporting literature evidence.

Existing approaches for templatization and retrieval of domain knowledge for SQL queries for the enterprise databases have the disadvantages that they do not understand the context of natural language query and retrieve the most suitable templatized domain statement for SQL query generation. Embodiments disclosed herein provide a method and system for templatization and retrieval of domain knowledge for enterprise text-to-SQL semantic parsing. In this approach, the system generates the templatized domain statements for the domain database, the templatized domain statements assist in understanding the natural language query and guides the model for generating the SQL query and retrieve the most suitable templatized domain statement for any natural language query to guide the trained model to generate the associated SQL query.

The embodiments of the present disclosure herein address unresolved problems of cyclic dependency between cost of power procurement and routing of one or more electric vehicles (EVs) of a logistic enterprise. Embodiments herein provide a method and system for an optimization framework to overcome a cyclic dependency between cost of power procurement and routing of electric vehicles (EVs) of a logistic enterprise. The optimization framework comprises two independent routines, coupled through an exchange of parameter values. First routine optimizes routing cost of the EVs to satisfy delivery constraints by assuming that the average EV charging cost is known. This is given as input to a second routine that optimizes electricity procurement cost from different energy sources by assuming that the EVs routes are fixed. The output from the second routine is then given back as input to first routine, and the procedure iterates till a stopping criteria is reached.

Existing model deployment approaches have the disadvantage that they do not consider feasibility of cloud instances for hosting a given LLM model. Embodiments disclosed herein provide a method and system for deployment of LLMs in a plurality of cloud instances. The system checks feasibility of the plurality of cloud instances for hosting an LLM, based on size of the LLM and storage space in each of the cloud instances. Further, a latency value for a plurality of batch sizes is determined for a plurality of LLM-accelerator pairs, in each of the plurality of cloud instances identified as feasible based on the feasibility check, using a performance model. Furthermore, a recommendation of one of the plurality of cloud instances identified as feasible is generated, based on the determined latency, a measured cost of deployment, a user workload, an application type, a plurality of latency constraints, and an evaluated performance.

Most of the existing testing tools automate the process of test case execution, however, these depend largely on an expert to manually write the test cases and pass parameters of the software to the test cases. Embodiments of present disclosure provide a method and system for automated parameterized software testing. The method involves- test case design, test case execution and report generation. The test case design involves identifying a set of parameters from a test automation suite for testing a web-based software application. Then, relevant test cases are identified, and a second set of parameters are obtained from the web-based software application. Next, test case execution is done through one or more of: i) a code development tool, ii) a command with arguments, iii) a build tool, and iv) an application server. Finally, a report is generated based on test case execution indicating the status of the test cases.

Existing multi-agent system are employed collaboratively, where each agent specialized in extracting entities, relations, or events from documents which is a more generalized collaborative framework without specific customizations. The disclosed method presents a collaborative multi-agent knowledge graph Retrieval-Augmented Generation (RAG) framework using a plurality of collaborative multi-agent LLMs for extracting information from a plurality of multimodal documents and constructing the unified knowledge graph to facilitate accurate and efficient query-based retrieval. A query generator agent generates a plurality of queries that aim to uncover all possible information present in the plurality of multimodal documents. A domain model is generated by a domain model generator agent based on the plurality of queries. The domain model populator agent populates the domain model with data extracted from the plurality of multimodal documents. A unified knowledge graph is constructed from ta plurality of populated domain models and organizes the extracted information for accurate retrieval.

This disclosure relates generally to a method of separating one or more target gases from a flue gas using fixed bed reactors. The target gas separation involves swing-adsorption mechanisms. State-of-the-art methods utilizing fixed beds suffers relatively high-pressure drops. To overcome the pressure-drop, coarser particles are used but this further lowers the adsorption capacity of the target gas. These drawbacks also affect the size of the fixed bed reactor, limiting its applicability for gas separation in mobile sources. The present invention provides the target gas separation from the fixed beds that involves simultaneous and sequential treatment of the flue gas in various stages of the swing-adsorption mechanisms. An integral heat recovery unit facilitates a plurality of transfers during target gas separation from the flue gas and manages entire heat requirement of the target gas separation process.

Multiple Instance Learning (MIL) is a useful method for extracting information from Gigapixel images using weakly supervised learning. Traditionally, it was approached with the hypothesis of considering each instance as an independent and identically distributed entity. Recently, it has been suggested that correlations between different instances would yield better results for classification. The present disclosure is designed to improve classification performance significantly. The self-selective MIL of the present disclosure facilitates binary classification with enhanced insights into more aggressive tumor regions.

This disclosure relates generally to a method and system for aesthetically placing and subconsciously nudging a plurality of advertisement products within a display screen of a mobile device. State-of-the-art methods demonstrate traditional way of marketing a single product being promoted by an influencer in a short video on social media platforms. The influencer captures a main frame of the display screen while advertising the product. However, most of the space on the display screen as well as the budget associated with an advertisement remain under-utilized. The present disclosure addresses these problems through a method of creating an overlay of a grid having a plurality of aesthetic spaces within the display screen. The aesthetic spaces are supplemented with an aural and an haptic experience via embedded audio files. The grid dynamically aligns the plurality of aesthetic spaces to eye-gaze hotspots to provide an individualistic personalization of the grid.

Proliferation of edge devices has significantly advanced technologies in sectors such as autonomous driving and surveillance. However, deploying machine learning models on these resource-constrained devices presents challenges including scalability and managing unpredictable workloads thereby affecting real-time performance in edge-only environments. The present disclosure discloses a method and system for workload deployment in a content guided and service level agreement (SLA) aware edge-cloud architecture. In the present disclosure, a camera feed content-guided load balancing technique is provided that dynamically manages workloads between edge and cloud. Features are extracted from an incoming camera feed to perform load-balancing process efficiently. The load balancer determines a maximum number of concurrent feeds for processing at the edge, with the remaining feeds handled by the cloud based on content of the incoming camera feed. Additionally, a cost model to estimate expenses of deploying the edge-cloud architecture in real-world scenarios with dynamic workloads is provided.

The disclosure relates generally to methods and systems for dynamic drug design of a pharmacological target. Conventional techniques in the drug design that use in-silico, in-vitro, and in-vivo approaches are not explicitly mentioned workflow related details for identifying novel molecules. The methods and systems of the present disclosure make the drug design dynamically by integrating the in-silico, in-vitro and in-vivo approaches through the dynamic generative artificial intelligence (GenAI) and artificial intelligence (AI) technologies. The integration of in-silico (molecular modeling and AI), in-vitro and in-vivo approaches helps in designing the novel optimized lead molecules. Optimization and prediction of ADMET based on QM based descriptors help in filtering the molecules.

Conventional models extract chemical data through querying tables by decomposing complex user queries. This disclosure relates generally to a method and system for extraction of chemical molecules and associated target properties from text and complex tables using Large Language Models (LLMs). The disclosed method extracts a plurality of molecular property values associated with the chemical molecules, from a plurality data sources, via generating a chemical composition schema utilizing LLMs. The chemical composition schema acts as a lens to view tabular information and a text. Relevant tables are identified and tabular information comprising chemical composition instances are extracted from a plurality of documents utilizing the chemical composition schema, along with LLMs and prompting techniques. The chemical composition instances are curated and reconciled into a unified knowledge graph, which is then used for querying. The disclosed method ensures precision in tabular information extraction without the need for extensive model training.

Current approaches for cross pollination contamination detection mainly rely on field surveys that are generally costly and consumes lot of time. The genetic purity of seed estimation techniques needs specialized equipment which again are associated with high cost and longer time. Present disclosure provides method and system for genetic purity estimation and cross-pollination detection in agricultural fields. The system receives remote sensing, weather, historical and soil health data of region of interest. The system then uses data to calculate contamination score representing contamination due to cross pollination. Thereafter, system uses data to determine field suitability and stress score. Further, system detects deviations in crop characteristics indicative of off-types in and around region of interest based on data. Finally, system estimates genetic purity of seeds coming from crop plotted in region of interest based on contamination, field suitability and field stress score, and off-types to determine genetic purity of seed score.

Zusammenfassung wird geladen …

Method and system disclosed uses LLM with multi lingual support to helps identify phishing emails. The method creates sufficient visual indication for the user to notice it as an unusual email message. The method illustrates the email using relevant illustrations and / or emoticons in the notification bar of the phone or the desktop or web email client. Changes the font type face with kerning models as per the device form factor. By doing this disguised domain names or misspelled words can easily be noticed by the user. A sentiment analysis is utilized to identify if the email message is creating a sense of urgency for purported negative or positive event for the user. The result of sentiment analysis is then used to also recommend 2 factor authentication. Thus allow user to establish credibility of the email. accordingly the email header and contents are updated appropriately.

Zusammenfassung wird geladen …

Existing techniques fail to propose a method for improving model parameters optimization and seamless integration of clinical data with computational model required for personalized cardiac hemodynamic model development. This disclosure relates to a system and method, which receives one or more model parameters comprising height, active pressure components and passive pressure components specific to left ventricle from cardiac hemodynamic model. The first set of values corresponding to one or more input parameters comprising end systolic diameter and end diastole diameter are received from subject specific Echo data. The second set of values corresponding to one or more model parameters are estimated using one or more input parameters. The estimated second set of values are optimized using a Particle swarm optimization to obtain one or more optimized values. Select at least a subset of optimized one or more values corresponding to one or more model parameters for personalization.

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …