Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Amazon IoT Greengrass Version 1 entered the extended life phase on June 30, 2023. For more information, see the Amazon IoT Greengrass V1 maintenance policy. After this date, Amazon IoT Greengrass V1 won't release updates that provide features, enhancements, bug fixes, or security patches. Devices that run on Amazon IoT Greengrass V1 won't be disrupted and will continue to operate and to connect to the cloud. We strongly recommend that you migrate to Amazon IoT Greengrass Version 2, which adds significant new features and support for additional platforms.

ML Feedback connector

The ML Feedback connector makes it easier to access your machine learning (ML) model data for model retraining and analysis. The connector:

To configure this connector, you describe your supported feedback configurations in JSON format. A feedback configuration defines properties such as the destination Amazon S3 bucket, content type, and sampling strategy. (A sampling strategy is used to determine which samples to upload.)

You can use the ML Feedback connector in the following scenarios:

ARN: arn:aws:greengrass:region::/connectors/MLFeedback/versions/1

Requirements

This connector has the following requirements:

Parameters

Create Connector Example (Amazon CLI)

The following CLI command creates a ConnectorDefinition with an initial version that contains the ML Feedback connector.

aws greengrass create-connector-definition --name MyGreengrassConnectors --initial-version '{
    "Connectors": [
        {
            "Id": "MyMLFeedbackConnector",
            "ConnectorArn": "arn:aws:greengrass:region::/connectors/MLFeedback/versions/1",
            "Parameters": {
                "FeedbackConfigurationMap": "{  \"RandomSamplingConfiguration\": {  \"s3-bucket-name\": \"my-aws-bucket-random-sampling\",  \"content-type\": \"image/png\",  \"file-ext\": \"png\",  \"sampling-strategy\": {  \"strategy-name\": \"RANDOM_SAMPLING\",  \"rate\": 0.5  } },  \"LeastConfidenceConfiguration\": {  \"s3-bucket-name\": \"my-aws-bucket-least-confidence-sampling\",  \"content-type\": \"image/png\",  \"file-ext\": \"png\",  \"sampling-strategy\": {  \"strategy-name\": \"LEAST_CONFIDENCE\",  \"threshold\": 0.4  } } }", 
                "RequestLimit": "10"
            }
        }
    ]
}'

FeedbackConfigurationMap example

The following is an expanded example value for the FeedbackConfigurationMap parameter. This example includes several feedback configurations that use different sampling strategies.

{
    "ConfigID1": {
        "s3-bucket-name": "my-aws-bucket-random-sampling",
        "content-type": "image/png",
        "file-ext": "png",
        "sampling-strategy": {
            "strategy-name": "RANDOM_SAMPLING",
            "rate": 0.5
        }
    },
    "ConfigID2": {
        "s3-bucket-name": "my-aws-bucket-margin-sampling",
        "content-type": "image/png",
        "file-ext": "png",
        "sampling-strategy": {
            "strategy-name": "MARGIN",
            "threshold": 0.4
        }
    },
    "ConfigID3": {
        "s3-bucket-name": "my-aws-bucket-least-confidence-sampling",
        "content-type": "image/png",
        "file-ext": "png",
        "sampling-strategy": {
            "strategy-name": "LEAST_CONFIDENCE",
            "threshold": 0.4
        }
    },
    "ConfigID4": {
        "s3-bucket-name": "my-aws-bucket-entropy-sampling",
        "content-type": "image/png",
        "file-ext": "png",
        "sampling-strategy": {
            "strategy-name": "ENTROPY",
            "threshold": 2
        }
    },
    "ConfigID5": {
        "s3-bucket-name": "my-aws-bucket-no-sampling",
        "s3-prefix": "DeviceA",
        "content-type": "application/json"
    }
}

Sampling strategies

The connector supports four sampling strategies that determine whether to upload samples that are passed to the connector. Samples are discrete instances of data that a model uses for a prediction. You can use sampling strategies to filter for the samples that are most likely to improve model accuracy.

Input data

User-defined Lambda functions use the publish function of the feedback client in the Amazon IoT Greengrass Machine Learning SDK to invoke the connector. For an example, see Usage Example.

The publish function takes the following arguments:

Output data

This connector publishes data to three MQTT topics:

You must configure subscriptions to allow the connector to communicate on MQTT topics. For more information, see Inputs and outputs.

Usage Example

The following example is a user-defined Lambda function that uses the Amazon IoT Greengrass Machine Learning SDK to send data to the ML Feedback connector.

import json
import logging
import os
import sys
import greengrass_machine_learning_sdk as ml

client = ml.client('feedback')

try:
    feedback_config_id = os.environ["FEEDBACK_CONFIG_ID"]
    model_input_data_dir = os.environ["MODEL_INPUT_DIR"]
    model_prediction_str = os.environ["MODEL_PREDICTIONS"]
    model_prediction = json.loads(model_prediction_str)
except Exception as e:
    logging.info("Failed to open environment variables. Failed with exception:{}".format(e))
    sys.exit(1)

try:
    with open(os.path.join(model_input_data_dir, os.listdir(model_input_data_dir)[0]), 'rb') as f:
        content = f.read()
except Exception as e:
    logging.info("Failed to open model input directory. Failed with exception:{}".format(e))
    sys.exit(1)    

def invoke_feedback_connector():
    logging.info("Invoking feedback connector.")
    try:
        client.publish(
            ConfigId=feedback_config_id,
            ModelInput=content,
            ModelPrediction=model_prediction
        )
    except Exception as e:
        logging.info("Exception raised when invoking feedback connector:{}".format(e))
        sys.exit(1)    

invoke_feedback_connector()

def function_handler(event, context):
    return

Licenses

The ML Feedback connector includes the following third-party software/licensing:

This connector is released under the Greengrass Core Software License Agreement.

See also