18048610. INTELLIGENT DEVICE DATA FILTER FOR MACHINE LEARNING simplified abstract (INTERNATIONAL BUSINESS MACHINES CORPORATION)

INTELLIGENT DEVICE DATA FILTER FOR MACHINE LEARNING

Organization Name

INTERNATIONAL BUSINESS MACHINES CORPORATION

Inventor(s)

Janice Yang of Cambridge MA (US)

Kate Hogan of St. Louis MO (US)

Arnav Aggarwal of Los Altos CA (US)

Ethan Weaver of Morgantown WV (US)

John S. Werner of Fishkill NY (US)

INTELLIGENT DEVICE DATA FILTER FOR MACHINE LEARNING - A simplified explanation of the abstract

This abstract first appeared for US patent application 18048610 titled 'INTELLIGENT DEVICE DATA FILTER FOR MACHINE LEARNING

Simplified Explanation

The patent application describes a method where a processor at an edge device running a local instance of a machine learning model receives a set of inference data, runs the data through filters to determine the probability of sending each data point back to the cloud environment for retraining the ML model.

• The processor receives inference data at an edge device running a local instance of a machine learning model.
• The machine learning model was trained in a cloud environment and then deployed to the edge device.
• The processor determines the probability for each data point of whether it should be sent back to the cloud environment for retraining.
• A send-back threshold is set to determine when a data point should be sent back for retraining.

Potential Applications

This technology could be applied in various industries such as healthcare, finance, and manufacturing for real-time data analysis and model retraining at the edge.

Problems Solved

1. Enables real-time decision-making at the edge without relying solely on cloud resources. 2. Improves the efficiency of machine learning models by retraining them with relevant data points.

Benefits

1. Reduces latency by processing data locally at the edge device. 2. Enhances the accuracy of machine learning models by continuously updating them with new data. 3. Optimizes resource utilization by sending only relevant data points back to the cloud for retraining.

Potential Commercial Applications

Optimizing Edge Machine Learning Model Retraining for Improved Performance

Possible Prior Art

Prior art may include research papers or patents related to edge computing, machine learning model deployment, and data retraining strategies.

Unanswered Questions

== How does this method handle data privacy and security concerns at the edge device? The patent application does not address the specific mechanisms in place to ensure data privacy and security at the edge device. This could be a potential area for further exploration and development.

== What are the potential limitations of using this method for continuous model retraining? The patent application does not discuss any limitations or challenges that may arise when continuously retraining machine learning models at the edge. Further research and testing may be needed to identify and address any potential drawbacks.

Original Abstract Submitted

In an approach, a processor receives, at an edge device running a local instance of a machine learning (ML) model, a set of inference data comprising a plurality of datapoints, wherein the local instance of the ML model is a deployed version of the ML model running in a cloud environment, and wherein the ML model was trained in the cloud environment and then deployed to the edge device. A processor runs the plurality of datapoints through one or more filters to determine a probability for each datapoint of whether a respective datapoint should be sent back to the cloud environment and used for retraining the ML model. A processor determines, for each datapoint, whether the probability for the respective datapoint meets a send back threshold that is required to be met before the respective datapoint is sent back to the cloud environment.