Samsung Electronics Co., Ltd.

Tata Consultancy Services Limited

Arun George

Sounak Dey

Arijit Mukherjee

Vishak Guddekoppa

Dighanchal Banerjee

Arpan Pal

Achanna Anil Kumar

Andrew Gigie

Roshan R Nair

Preetham Parshwanath Jain

Parag Jain

Dinesh Ratra

Rohit Kailash Sharma

Shivam

Tapas Chakravarty

Arijit Chowdhury

Smriti Rani

Rajesh Krishnan

Anshul Sharma

Kuchibhotla Aditi

Vishak G

Rakesh Nadig

Mitesh Sanjay Mutha

A method for identifying a plurality of workloads in a heterogenous environment includes collecting a plurality of parameters from at least one layer of a system stack associated with the plurality of workloads, correlating the collected plurality of parameters from different layers of software stack, and creating a feature set based on the correlated plurality of parameters. The method further includes processing the feature set using a successively ordered classifier chain (SOCC) module to identify the presence of the plurality of workloads in the heterogenous environment in a data center.

Methods and systems for identifying multiple workloads in a heterogeneous environment

This disclosure relates generally to time series forecasting, and, more particularly, to a system and method for online time series forecasting using spiking reservoir. Existing systems do not cater for efficient online time-series analysis and forecasting due to their memory and computation power requirements. System and method of the present disclosure convert a time series value F(t) at time 't' to an encoded multivariate spike train and extracts temporal features from the encoded multivariate spike train by the excitatory neurons of a reservoir, predict a time series value Y(t+k) at time ‘t’ by performing a linear combination of extracted temporal features with read-out weights, compute an error for predicted time series value Y(t+k) with input time series value F(t+k), employs a FORCE learning on read-out weights using the error to reduce error in future forecasting. Feeding a feedback value back to the reservoir to optimize memory of the reservoir.

System and method for online time-series forecasting using spiking reservoir

The disclosure provides a method, a distributed controller and a system for managing sequential storage devices in a distributed storage environment. The method comprising receiving a client allocation request from a client device and determining at least one zone set from a plurality of zone set based on the client allocation request and predefined I/O parameters. Each zone set of the plurality of zone set comprises a plurality of zones of sequential storage devices. Thereafter, method comprises provisioning the at least one zone set to the client device based on the client allocation request, thereby, managing the sequential storage devices. Each zone of a sequential storage device is arranged into a zone set based on at least one of sequential storage device parameters, and zone parameters. The use of the distributed controller allows optimizing the performance of the sequential storage devices and optimizing the network bandwidth utilization.

Method, distributed controller, and system for managing sequential storage devices in distributed storage environment

This disclosure relates generally to action recognition and more particularly to system and method for real-time radar-based action recognition. The classical machine learning techniques used for learning and inferring human actions from radar images are compute intensive, and require volumes of training data, making them unsuitable for deployment on network edge. The disclosed system utilizes neuromorphic computing and Spiking Neural Networks (SNN) to learn human actions from radar data captured by radar sensor(s). In an embodiment, the disclosed system includes a SNN model having a data pre-processing layer, Convolutional SNN layers and a Classifier layer. The preprocessing layer receives radar data including doppler frequencies reflected from the target and determines a binarized matrix. The CSNN layers extracts features (spatial and temporal) associated with the target's actions based on the binarized matrix. The classifier layer identifies a type of the action performed by the target based on the features.

System and method for real-time radar-based action recognition using spiking neural network(SNN)

Gesture recognition is a key requirement for Human Computer Interaction (HCl) and multiple modalities are explored in literature. Conventionally, channel taps are estimated using least square based estimation and tap corresponding to finger motion is tracked. These assume that noise component is negligible and can reduce the tracking accuracy for low SNR. Thus, to mitigate the above-mentioned limitation, the system and method of the present disclosure explore the feasibility of using speaker and microphone setup available in most of smart devices and transmit inaudible frequencies (acoustic) for detecting the human finger level gestures accurately. More specifically, System implements the method for millimeter level finger tracking and low power gesture detection on this tracked gesture. The system uses a subspace based high resolution technique for delay estimation and use microphone pairs to jointly estimate the multi-coordinates of finger movement.

System and method for acoustic based gesture tracking and recognition using spiking neural network

This disclosure generally relates optimized spike encoding for spiking neural networks (SNNs). The SNN processes data in spike train format, whereas the real world measurements/input signals are in analog (continuous or discrete) signal format; therefore, it is necessary to convert the input signal to a spike train format before feeding the input signal to the SNNs. One of the challenges during conversion of the input signal to the spike train format is to ensure retention of maximum information between the input signal to the spike train format. The disclosure reveals an optimized encoding method to convert the input signal to optimized spike train for spiking neural networks. The disclosed optimized encoding approach enables maximizing mutual information between the input signal and optimized spike train by introducing an optimal Gaussian noise that augments the entire input signal data.

Method and system for optimized spike encoding for spiking neural networks

A method of managing sequential storage devices in a distributed storage environment including receiving, by a distributed controller, requests for performing operations, from one or more client devices in a distributed storage environment, each request including data of a client device from the one or more client devices and an identifier of a host device from one or more host devices, each of the one or more host devices being associated with one or more sequential storage devices, determining, by the distributed controller, a sequential order of the requests, based on the identifiers included in the requests, and performing, by the distributed controller in the sequential order, the operations on the one or more sequential storage devices.

A method for dynamically replicating data in a disaggregated distributed storage system includes receiving, by a processor, a request from a controller to replicate data; dynamically selecting, by the processor, a storage node for replicating the data from among a plurality of storage nodes, based on one or more replication parameters associated with each storage node of the plurality of storage nodes, wherein the one or more replication parameters include at least one of a flash factor corresponding to the each storage node, and a latency between the controller and the each storage node; and providing, by the processor, information about the selected storage node to the controller for replicating the data.

Device and method for data replication in disaggregated distributed storage system

A method of performing replication recovery operation in a disaggregated distributed storage system (DDSS) includes identifying an affected storage node included in the storage pool based on an occurrence of one or more predetermined conditions; selecting at least one replacement storage node from among one or more storage nodes included in the storage pool to recover replication data stored in the affected storage node; selecting at least one controller for performing a replication recovery operation from among one or more controllers included in the controller pool; and assigning one or more replication recovery jobs to the at least one controller; and receiving, by the replication recovery device, a notification from the at least one controller based on the one or more replication recovery jobs being completed.

Method and system for performing replication recovery operation in a disaggregated distributed storage systems

Conventional gesture detection approaches demand large memory and computation power to run efficiently, thus limiting their use in power and memory constrained edge devices. Present application/disclosure provides a Spiking Neural Network based system which is a robust low power edge compatible ultrasound-based gesture detection system. The system uses a plurality of speakers and microphones that mimics a Multi Input Multi Output (MIMO) setup thus providing requisite diversity to effectively address fading. The system also makes use of distinctive Channel Impulse Response (CIR) estimated by imposing sparsity prior for robust gesture detection. A multi-layer Convolutional Neural Network (CNN) has been trained on these distinctive CIR images and the trained CNN model is converted into an equivalent Spiking Neural Network (SNN) via an ANN (Artificial Neural Network)-to-SNN conversion mechanism. The SNN is further configured to detect/classify gestures performed by user(s).

Acoustic system and method based gesture detection using spiking neural networks

A method for converting an electronic flash storage device having a byte addressable storage (ByAS) and a block addressable flash storage (BlAS) to a single byte addressable storage includes receiving, by a host, a request for memory allocation from the ByAS, the receiving being from a first application among of a plurality of applications running on a processor; deallocating, by the host, a least relevant page allocated to at least one second application among the plurality of applications; moving, by the host, a content to the BlAS at a first BlAS location, the content related to the least relevant page, the moving based on the deallocation; allocating, by the host, the least relevant page to the first application; and updating, by the host, a cache metadata and a page lookup table of the first application and the at least one second application based on the deallocation and allocation.

Method and system for converting electronic flash storage device to byte-addressable nonvolatile memory module

This disclosure relates to method of identifying a gesture from a plurality of gestures using a reservoir based convolutional spiking neural network. A two-dimensional spike streams is received from neuromorphic event camera as an input. The two-dimensional spike streams associated with at least one gestures from a plurality of gestures is preprocessed to obtain plurality of spike frames. The plurality of spike frames is processed by a multi layered convolutional spiking neural network to learn plurality of spatial features from the at least one gesture. A filter block is deactivated from the plurality of filter blocks corresponds to at least one gesture which are not currently being learnt. A spatio-temporal features is obtained by allowing the spike activations from CSNN layer to flow through the reservoir. The spatial feature is classified by classifier from the CSNN layer and the spatio-temporal features from the reservoir to obtain set of prioritized gestures.

System and method of gesture recognition using a reservoir based convolutional spiking neural network

A method for flash-aware heap memory management includes reserving a contiguous virtual space in a memory space of at least one process with a size equivalent to a size of a flash-based byte addressable device. The method also includes partitioning by a host device the memory space of the flash-based byte addressable device into multiple chunks. Each chunk includes multiple logical segments. The host device receives a memory allocation request from a thread associated with an application. The host device determines at least one chunk from the multiple chunks, including a least free logical segment compared to the other chunks from the multiple chunks. The host device allocates to the thread at least one chunk that includes the least free logical segment.

Method and system for flash-aware heap memory management wherein responsive to a page fault, mapping a physical page (of a logical segment) that was previously reserved in response to another page fault for another page in the first logical segment

Growing community of inventors

Bengaluru, India

Arun George