18274531. DISTRIBUTED MACHINE LEARNING WITH NEW LABELS USING HETEROGENEOUS LABEL DISTRIBUTION simplified abstract (Telefonaktiebolaget LM Ericsson (publ))

DISTRIBUTED MACHINE LEARNING WITH NEW LABELS USING HETEROGENEOUS LABEL DISTRIBUTION

Organization Name

Telefonaktiebolaget LM Ericsson (publ)

Inventor(s)

Gudur Gautham Krishna of Chennai (IN)

Satheesh Kumar Perepu of Chennai (IN)

DISTRIBUTED MACHINE LEARNING WITH NEW LABELS USING HETEROGENEOUS LABEL DISTRIBUTION - A simplified explanation of the abstract

This abstract first appeared for US patent application 18274531 titled 'DISTRIBUTED MACHINE LEARNING WITH NEW LABELS USING HETEROGENEOUS LABEL DISTRIBUTION

Simplified Explanation

The abstract describes a method for distributed machine learning where a first dataset with labels is provided to multiple local computing devices. The devices train local ML models and send back model probabilities values. A weights matrix is generated using these values, and a new set of model probabilities values is generated through sampling.

• Explanation of the patent:
• Multiple local computing devices are used for distributed machine learning.
• Local ML models are trained on a first dataset with labels.
• Model probabilities values are sent back to a central server.
• A weights matrix is generated using these values.
• New model probabilities values are generated through sampling.

Potential Applications

The technology described in this patent application could be applied in various fields such as:

• Healthcare for analyzing medical data distributed across different hospitals.
• Financial services for fraud detection using data from multiple sources.
• Autonomous vehicles for processing sensor data from different vehicles.

Problems Solved

The method described in the patent application addresses the following issues:

• Efficiently training machine learning models on distributed datasets.
• Combining model probabilities values from different sources.
• Generating new model probabilities values through sampling.

Benefits

The technology offers the following benefits:

• Improved accuracy of machine learning models through collaboration.
• Scalability for handling large datasets distributed across multiple devices.
• Privacy preservation by keeping data local and only sharing model probabilities values.

Potential Commercial Applications

A potential commercial application of this technology could be in:

• Cloud computing services for offering distributed machine learning capabilities.
• Data analytics platforms for processing data from various sources.
• Cybersecurity solutions for detecting anomalies in network traffic.

Possible Prior Art

One possible prior art for this technology could be:

• Research papers on federated learning techniques for collaborative model training.

How does the method handle discrepancies in model probabilities values from different devices?

The method does not specify how discrepancies in model probabilities values are resolved or accounted for. This could impact the accuracy of the final model.

What is the computational overhead of generating and using the weights matrix in the distributed machine learning process?

The patent application does not provide information on the computational resources required for generating and utilizing the weights matrix. This could be a crucial factor in determining the feasibility of the method in practical applications.

Original Abstract Submitted

A method for distributed machine learning (ML) which includes providing a first dataset including a first set of labels to a plurality of local computing devices including a first local computing device and a second local computing device. The method further includes receiving, from the first local computing device, a first set of ML model probabilities values from training a first local ML model using the first set of labels. The method further includes receiving, from the second local computing device, a second set of ML model probabilities values from training a second local ML model using the first set of labels and one or more labels different from any label in the first set of labels. The method further includes generating a weights matrix using the received first set of ML model probabilities values and the received second set of MIL model probabilities values. The method further includes generating a third set of ML model probabilities values by sampling using the generated weights matrix.