Tata Consultancy Services Limited

Pankaj Malhotra

Gautam Shroff

Lovekesh Vig

Puneet Kumar Agarwal

Priyanka Gupta

Venkataramana Runkana

Narendhar Gugulothu

Jyoti Narwariya

Diksha Garg

Aditya Pareek

Vishnu Swaroopji Masampally

Sivakumar Subramanian

Anirudh Deodhar

Ashit Gupta

Ankit Sharma

Muralikrishnan Ramanujam

Mukul Patil

Solomon Pushparaj Manuelraj

Vishnu Tv

Charan Theja Dhanda

Dhandapani Subramaniam

Lalith Roshanlal Jain

Joel Thomson Diraviam Andrew

Sai Prasad Parameswran

Vishnu Tankasala Veparala

Vibhor Gupta

Nishit Raj

Ankit Bansal

Vishal Jadhav

Vishnu T V

Gupta Priyanka

Vig Lovekesh

Shroff Gautam

Recommender Systems (RS) tend to recommend more popular items instead of the relevant long-tail items. Mitigating such popularity bias is crucial to ensure that less popular but relevant items are recommended. System described herein analyses popularity bias in session-based RS obtained via deep learning (DL) models. DL models trained on historical user-item interactions in session logs (having long-tailed item-click distributions) tend to amplify popularity bias. To understand source of this bias amplification, potential sources of bias at data-generation stage (user-item interactions captured as session logs) and model training stage are considered by the system for recommendation wherein popularity of item has causal effect on user-item interactions via conformity bias, and item ranking from models via biased training process due to class imbalance. While most existing approaches address only one of these effects, a comprehensive causal inference framework is implemented by present disclosure that identifies and mitigates effects at both stages.

System, method, and non-transitory machine-readable information storage medium for recommendations of items and controlling an associated bias thereof

Session-based Recommendation (SR) is the task of recommending the next item based on previously recorded user interactions. However, most existing approaches for SR either rely on costly online interactions with real users (model-free approaches) or rely on potentially biased rule-based or data-driven user-behavior models (model-based approaches) for learning. This disclosure relates to a system and method for selecting session-based recommendation policies using historical recommendations and user feedback. Herein, the learning of recommendation policies given offline or batch data from old recommendation policies based on a Distributional Reinforcement Learning (DRL) based recommender system in the offline or batch-constrained setting without requiring access to a user-behavior model or real-interactions with the users.

System and method for training recommendation policies

Several applications capture data from sensors resulting in multi-sensor time series. Existing neural networks-based approaches for such multi-sensor/multivariate time series modeling assume fixed input-dimension/number of sensors. Such approaches can struggle in practical setting where different instances of same device/equipment come with different combinations of installed sensors. In the present disclosure, neural network models are trained from such multi-sensor time series having varying input dimensionality, owing to availability/installation of different sensors subset at each source of time series. Neural network (NN) architecture is provided for zero-shot transfer learning allowing robust inference for multivariate time series with previously unseen combination of available dimensions/sensors at test time. Such combinatorial generalization is achieved by conditioning layers of core NN-based time series model with 'conditioning vector' carrying information of available sensors combination for each time series and is obtained by summarizing learned “sensor embedding vectors set” corresponding to available sensors in time series.

Neural networks for handling variable-dimensional time series data

This disclosure relates generally to method and system for handling popularity bias in item recommendations. In an embodiment the method includes initializing an item embedding look-up matrix corresponding to items in a sequence of item-clicks with respect to a training data. L2 norm is applied to the item embedding look-up matrix to learn a normalized item embeddings. Using a neural network, a session embeddings corresponding to the sequences of item-clicks is modeled and L2 norm is applied to the session embeddings to obtain a normalized session embeddings. Relevance scores corresponding to each of the plurality of items are obtained based on similarity between the normalized item embeddings and the normalized session embeddings. A multi-dimensional probability vector corresponding to the relevance scores for the items to be clicked in the sequence is obtained. A list of the items ordered based on the multi-dimensional probability vector is provided as recommendation.

System, method, and non-transitory machine readable information storage medium for handling popularity bias in item recommendations

This disclosure relates generally to a method and system for real time monitoring and forecasting of fouling of an air preheater (APH) in a thermal power plant. The system is deploying a digital replica or digital twin that works in tandem with the real APH of the thermal power plant. The system receives real-time data from one or more sources and provides real-time soft sensing of intrinsic parameters as well as that of health, fouling related parameters of APH. The system is also configured to diagnose the current class of fouling regime and the reasons behind a specific class of fouling regime of the APH. The system is also configured to be used as advisory system that alerts and recommends corrective actions in terms of either APH parameters or parameters controlled through other equipment such as selective catalytic reduction or boiler or changes in operation or design.

Method and system for realtime monitoring and forecasting of fouling of air preheater equipment

This disclosure relates to optimizing an operation of an equipment by a neural network based optimizer is provided. The method include receiving, information associated with at least one equipment instance (j) as an input at a predefined sequence of timestamps; training, a plurality of simulation models for each equipment instance to obtain a function (f); processing, the external input parameters (e) to obtain a fixed-dimensional vector and passed as an input to obtain an vector (i); generating, a modified (i) from the output vector (i) based on a domain constraint value; computing, a reward (r) based on (i) the function (f), (ii) the modified (i), (iii) the external input parameters (e), and (iv) a reward function (R); and iteratively performing the steps of processing, generating, and computing reward (r) for a series of subsequent equipment instances after expiry of the predefined sequence of timestamps associated with a first equipment instance.

System and method for operation optimization of an equipment

Neural networks can be used for time series data classification. However, in a K-shot scenario in which sufficient training data is unavailable to train the neural network, the neural network may not produce desired results. Disclosed herein are a method and system for training a neural network for time series data classification. In this method, by processing a plurality of task specific data, a system generates a set of updated parameters, which is further used to train a neural network (network) till a triplet loss is below a threshold. The network is trained on a diverse set of few-shot tasks sampled from various domains (e.g. healthcare, activity recognition, and so on) such that it can solve a target task from another domain using only a small number of training samples from the target task.

Method and system for training a neural network for time series data classification

Estimating Remaining Useful Life (RUL) from multi-sensor time series data is difficult through manual inspection. Current machine learning and data analytics methods, for RUL estimation require large number of failed instances for training, which are rarely available in practice, and these methods cannot use information from currently operational censored instances since their failure time is unknown. Embodiments of the present disclosure provide systems and methods for estimating RUL using time series data by implementing an LSTM-RNN based ordinal regression technique, wherein during training RUL value of failed instance(s) is encoded into a vector which is given as a target to the model. Unlike a failed instance, the exact RUL for a censored instance is unknown. For using the censored instances, target vectors are generated and the objective function is modified for training wherein the trained LSTM-RNN based ordinal regression is applied on an input test time series for RUL estimation.

Failed and censored instances based remaining useful life (RUL) estimation of entities

Anomaly detection from time series is one of the key components in automated monitoring of one or more entities. Domain-driven sensor selection for anomaly detection is restricted by knowledge of important sensors to capture only a certain set of anomalies from the entire set of possible anomalies. Hence, existing anomaly detection approaches are not very effective for multi-dimensional time series. Embodiments of the present disclosure depict sparse neural network for anomaly detection in multi-dimensional time series (MDTS) corresponding to a plurality of parameters of entities. A reduced-dimensional time series is obtained from the MDTS via an at least one feedforward layer by using a dimensionality reduction model. The dimensionality reduction model and recurrent neural network (RNN) encoder-decoder model are simultaneously learned to obtain a multi-layered sparse neural network. A plurality of error vectors corresponding to at least one time instance of the MDTS is computed to obtain an anomaly score.

Sparse neural network based anomaly detection in multi-dimensional time series

Traditional systems and methods have implemented hand-crafted feature extraction from varying length time series that results in complexity and requires domain knowledge. Building classification models requires large labeled data and is computationally expensive. Embodiments of the present disclosure implement learning models for classification tasks in multi-dimensional time series by performing feature extraction from entity's parameters via unsupervised encoder and build a non-temporal linear classifier model. A fixed-dimensional feature vector is outputted using a pre-trained unsupervised encoder, which acts as off-the shelf feature extractor. Extracted features are concatenated to learn a non-temporal linear classification model and weight is assigned to each extracted feature during learning which helps to determine relevant parameters for each class. Mapping from parameters to target class is considered while constraining the linear model to use only subset of large number of features.

Systems and methods for classification of multi-dimensional time series of parameters

This disclosure relates generally to health monitoring of systems, and more particularly to monitor health of a system for fault signature identification. The system estimates Health Index (HI) of the system as time series data. By analyzing data corresponding to the estimated HI, the system identifies one or more time windows in which majority of the estimated HI values are low as a low HI window, and one or more time windows in which majority of the estimated HI values are high as a high HI window. Upon identifying a low HI window, which indicates an abnormal behavior of the system being monitored, based on a local Bayesian Network generated for the system being monitored, an Explainability Index (EI) for each sensor is generated, wherein the EI quantifies contribution of the sensor to the low HI. Further, associated component(s) is identified as contributing to abnormal/faulty behavior of the system.

Method and system for health monitoring and fault signature identification

The present subject matter relates to entity resolution, and in particular, relates to providing an entity resolution from documents. The method comprises obtaining a plurality of documents corresponding to a plurality of entities, from at least one data source. Upon receiving the plurality of documents, the plurality of documents is blocked into at least one bucket based on textual similarity. Further, a graph including a plurality of record vertices and at least one bucket vertex is created. The plurality of record vertices and the at least one bucket vertex are indicative of the plurality of documents and the at least one bucket, respectively. Subsequently, a notification is provided to a user for selecting one of a Bucket-Centric Parallelization (BCP) technique and a Record-Centric Parallelization (RCP) technique for resolving entities from the plurality of documents. Based on the selection, a resolved entity-document for each entity is created.

Entity resolution from documents

The present subject matter relates to entity resolution, and in particular, relates to providing an entity resolution from documents. The method comprises obtaining the plurality of documents from at least one data source. The plurality of documents is blocked into at least one bucket based on textual similarity and inter-document references among the plurality of documents. Further, within each bucket, a merged document for each entity may be created based on an iterative match-merge technique. The iterative match-merge technique identifies, from the plurality of documents, at least one matching pair of documents and merges the at least one matching pair of documents to create the merged document for each entity. The merged documents may be merged to generate a resolved entity-document for each entity based on a graph clustering technique.

An anomaly detection system and method is provided. The system comprising: a hardware processor; and a memory storing instructions to configure the hardware processor, wherein the hardware processor receives a first time-series data comprising a first set of points and a second time-series data comprising a second set of points, computes a first set of error vectors for each point of the first set, and a second set of error vectors for each point of the second set, each set of error vectors comprising one or more prediction errors; estimates parameters based on the first set of error vectors comprising; applies (or uses) the parameters on the second set of error vectors; and detects an anomaly in the second time-series data when the parameters are applied on the second set of error vectors.

Growing community of inventors

Noida, India

Pankaj Malhotra