Hewlett Packard Enterprise Development LP

Krishnaprasad Lingadahalli Shastry

Sathyanarayanan Manamohan

Vishesh Garg

Satish Kumar Mopur

Saikat Mukherjee

Gunalan Perumal Vijayan

Ravi Sarveswara

Sridhar Balachandriah

Ashutosh Agrawal

Gregory S Battas

Avinash Chandra Pandey

Shounak Bandopadhyay

Sathyanarayana Manamohan

Systems and methods are provided for implementing a Siamese neural network using improved 'sub' neural networks and loss function. For example, the system can detect a granular change in images using a Siamese Neural Network with Convolutional Autoencoders as the twin sub networks (e.g., Siamese AutoEncoder or “SAE”). In some examples, the loss function may be an adaptive loss function to the SAE network rather than a contrastive loss function, which can help enable smooth control of granularity of change detection across the images. In some examples, an image separation distance value may be calculated to determine the value of change between the image pairs. The image separation distance value may be determined using an Euclidean distance associated with a latent space of an encoder portion of the autoencoder of the neural networks.

Enabling feature importance using Siamese autoencoders for effective image change detection

Systems and methods are disclosed for providing decentralized policy-based transactional object management for systems employing federated workflows. Various disclosed components may be added to one or more nodes of a decentralized network, wherein the disclosed components perform registration, replication, and read/write access interfacing functions. These functions result in the storage of objects on the decentralized network in a way which allows for decentralized, policy-based, and transactional management of the objects.

Decentralized policy-based transactional object management for federated workflows

A system and a method for adaptively synchronizing learning of multiple learning models are disclosed. Several local learning models are executed on multiple nodes. Learning model parameters are shared by such nodes to a master node, in multiple iterations, after a predefined synchronization interval. Such learning model parameters are aggregated and central learning models are generated based on aggregated set of learning model parameters. Accuracies of the central learning models and an average accuracy of the central learning models are determined. Accuracy of an immediate central learning model i.e. the one received after determining the average accuracy, is compared with the average accuracy. Based on the difference between the accuracy of the immediate central learning model and the average accuracy, the synchronization interval is modified, and the multiple nodes are updated about this modified synchronization interval.

Adaptively synchronizing learning of multiple learning models

Bias in Machine Learning (ML) is when an ML algorithm tends to incompletely learn relevant and important patterns from a dataset, or learns the patterns from data incorrectly. Such inaccuracy can cause the algorithm to miss important relationships between patterns and features in data, resulting in inaccurate algorithm predictions. Systems and methods for detecting potential ML bias in input image datasets are described herein. After a target image is received, a subset of images related to the target image is extracted. The target image and subset of images are analyzed under an imbalance assessment and data bias assessment to determine the presence of any potential data bias in a ML training pipeline. If any data bias is determined, one or more messages summarizing the assessments and including explanations to enable more accurate predictions in image assessments are sent to the user.

Image data bias detection with explainability in machine learning

Systems and methods are provided for retraining machine learning (ML) models. Examples may automatically identify skewed, anomalous, and/or drift occurrence data in real-world input data. By automatically identifying such data, examples can reduce subjectivity in ML model retraining as well as reduce time spent determining a need to retrain a ML model. Accordingly, a determination can be made objectively by a computing system or device according to computer-implemented instructions. Additionally, examples may automatically isolate and transfer data relevant to the retraining of a ML model to a training environment for retraining the ML model using real-world input data. Examples also synthesize large samples of data for use in retraining a ML model. The synthesized data may be generated based on the isolated and transferred data and can be used in place of actual real-world input data to reduce a corresponding delay.

Image data synthesis using adversarial autoencoders for continual training of models

Decentralized machine learning to build models is performed at nodes where local training datasets are generated. A blockchain platform may be used to coordinate decentralized machine learning (ML) over a series of iterations. For each iteration, a distributed ledger may be used to coordinate the nodes communicating via a blockchain network. A node can include self-healing features to recover from a fault condition within the blockchain network in manner that does not negatively impact the overall learning ability of the decentralized ML system. During self-healing, the node can determine that a local ML state is not consistent with the global ML state and trigger a corrective action to recover the local ML state. Thereafter, the node can generate a blockchain transaction indicating that it is in-sync with the most recent iteration of training, and informing other nodes to reintegrate the node into ML.

System and method for self-healing in decentralized model building for machine learning using blockchain

Systems and methods are provided for implementing swarm learning while using blockchain technology and election/voting mechanisms to ensure data privacy. Nodes may train local instances of a machine learning model using local data, from which parameters are derived or extracted. Those parameters may be encrypted and persisted until a merge leader is elected that can merge the parameters using a public key generated by an external key manager. A decryptor that is not the merge leader can be elected to decrypt the merged parameter using a corresponding private key, and the decrypted merged parameter can then be shared amongst the nodes, and applied to their local models. This process can be repeated until a desired level of learning has been achieved. The public and private keys are never revealed to the same node, and may be permanently discarded after use to further ensure privacy.

Secure parameter merging using homomorphic encryption for swarm learning

The disclosure relates to decentralized management of nodes in a blockchain network. Participants may agree to a consensus rules and implement them as smart contracts. For example, one rule may specify that a node will accept a change proposal only when its local policies and/or data allow it to implement the change. A smart contract may implement this rule and deploy it across the blockchain network for each node to follow. Other participants, through their nodes, may propose changes to the blockchain network, and each node may consult its copy of the smart contract to determine whether to vote to approve the change request and apply the change request locally.

System and method of decentralized management of multi-owner nodes using blockchain

Systems and methods are provided for leveraging blockchain technology in a swarm learning context, where nodes of a blockchain network that contribute data to training a machine learning model using their own local data can be rewarded. In order to conduct such data monetization in a fair and accurate manner, the systems and methods rely on various phases in which Merkle trees are used and corresponding Merkle roots are registered in a blockchain ledger. Moreover, any claims for a reward are challenged by peer nodes before the reward is distributed.

Systems and methods for monetizing data in decentralized model building for machine learning using a blockchain

Decentralized machine learning to build models is performed at nodes where local training datasets are generated. A blockchain platform may be used to coordinate decentralized machine learning over a series of iterations. For each iteration, a distributed ledger may be used to coordinate the nodes. Rules in the form of smart contracts may enforce node participation in an iteration of model building and parameter sharing, as well as provide logic for electing a node that serves as a master node for the iteration. The master node obtains model parameters from the nodes and generates final parameters based on the obtained parameters. The master node may write its state to the distributed ledger indicating that the final parameters are available. Each node, via its copy of the distributed ledger, may discover the master node's state and obtain and apply the final parameters to its local model, thereby learning from other nodes.

System and method of decentralized machine learning using blockchain

A system and method for accounting for the impact of concept drift in selecting machine learning training methods to address the identified impact. Pattern recognition is performed on performance metrics of a deployed production model in an Internet-of-Things (IoT) environment to determine the impact that concept drift (data drift) has had on prediction performance. This concurrent analysis is utilized to select one or more approaches for training machine learning models, thereby accounting for the temporal dynamics of concept drift (and its subsequent impact on prediction performance) in a faster and more efficient manner.

Systems and methods for determining machine learning training approaches based on identified impacts of one or more types of concept drift

The disclosure relates to a framework for dynamic management of analytic functions such as data processors and machine learned ('ML') models for an Internet of Things intelligent edge that addresses management of the lifecycle of the analytic functions from creation to execution, in production. The end user will be seamlessly able to check in an analytic function, version it, deploy it, evaluate model performance and deploy refined versions into the data flows at the edge or core dynamically for existing and new end points. The framework comprises a hypergraph-based model as a foundation, and may use a microservices architecture with the ML infrastructure and models deployed as containerized microservices.

Intelligent lifecycle management of analytic functions for an IoT intelligent edge with a hypergraph-based approach

The disclosure relates to technology that implements flow control for machine learning on data such as Internet of Things ('IoT') datasets. The system may route outputs of a data splitter function performed on the IoT datasets to a designated target model based on a user specification for routing the outputs. In this manner, the IoT datasets may be dynamically routed to target datasets without reprogramming machine-learning pipelines, which enable rapid training, testing and validation of ML models as well as an ability to concurrently train, validate, and execute ML models.

Intelligent IOT data split management

The disclosure relates to decentralized management of edge nodes operating outside an enterprise network using blockchain technology. A management node may operate within a firewall of the enterprise to manage the edge nodes operating outside the firewall using blockchain technology. The management node may coordinate management by writing change requests to a decentralized ledger. The edge nodes may read the change requests from its local copy of the distributed ledger and implement the change requests. Upon implementation, an edge node may broadcast its status to the blockchain network. The management node may mine the transactions from the edge nodes into the distributed ledger, thereby creating a secure and scalable way to coordinate management and record the current and historical system state. The system also provides the edge nodes with a cryptographically secured, machine-to-machine maintained, single version of truth, enabling them to take globally valid decision based on local data.

Growing community of inventors

Bengaluru, India

Krishnaprasad Lingadahalli Shastry