Amazon Technologies, Inc.

Aran Khanna

Calvin Yue-Ren Kuo

Sunil Mallya Kasaragod

Edge devices of a network collect data. An edge device may determine whether to process the data using a local data processing model or to send the data to a tier device. The tier device may receive the data from the edge device and determine whether to process the data using a higher tier data processing model of the tier device. If the tier device determines to process the data, then the tier device processes the data using the higher tier data processing model, generates a result based on the processing, and sends the result to an endpoint (e.g., back to the edge device, to another tier device, or to a control device). If the tier device determines not to process the data, then the tier device may send the data on to another tier device for processing by another higher tier model.

Model tiering for IoT device clusters

Techniques for model reutilization with heterogeneous sensor stacks via sensor data auto-normalization are described. A normalization model can be trained and utilized to normalize sensor data generated by a first type of sensor stack so that it can be used with an existing machine learning model that was trained using data from another type or types of sensor stacks having different characteristics. A sensor data can be generated by the sensor stack and provided as an input to the normalization model to yield normalized sensor data. The normalized sensor data can be provided as input to the existing model to generate accurate results despite the sensor stack having different characteristics than the one(s) used to train the machine learning model.

Model reutilization with heterogeneous sensor stacks via sensor data auto-normalization

A hub device of a network receives data from edge devices and generates a local result. The hub device also sends the data to a remote provider network and receives a result from the remote provider network, wherein the result is based on the data received from the edge devices. The hub device then generates a response based on the local result or the received result. The hub device may determine to correct the local result based on the result received from the remote provider network, and generate the response based on the corrected result. The hub device may generate an initial response before receiving the result from the provider network. For example, the hub device may determine that the confidence level for the local result is above the threshold level and in response, generate the initial response based on the local result.

Split predictions for IoT devices

A hub device of a network receives topology data for the network that indicates changes in a topology or performance of the network. The hub device determines modifications to processing functions for edge devices of the network based on the topology data. The hub device deploys the modifications to respective edge devices of the network. An edge device may collect data and use a processing function to perform operations on the data, generate processed data, and send the processed data to the hub device for further processing. In some cases, a remote provider network receives topology data for the network and generates modifications to processing functions for edge devices of the network based on the topology data. The remote provider network then transmits the modifications to the network.

Generating adaptive models for IoT networks

A model training service of a provider network receives data from edge devices of a remote network. The model training service analyzes the received data. The model training service may also analyze global data from other edge devices of other remote networks. The model training service may then generate updates to local data processing models based on the analysis. The updates are configured to update the local data processing models at the edge devices of the remote network. The provider network deploys the updates to the remote network. The updates are then applied to the data processing models of the edge devices.

Training models for IOT devices

A hub device of a network receives a data model that includes a secure portion that is encrypted and one or more unsecure portions. The hub device deploys the one or more unsecure portions of the data model to respective edge devices of the network. The hub device decrypts the secure portion of the data model. The edge devices collect data (e.g., from sensors) and process the data using the unsecure portions of the data model. The edge devices send the processed data to the hub device. The hub device performs operations on the received processed data using the decrypted secure portion of the data model in a secure execution environment (e.g., a TPM or other secure module). The secure portion of the data model generates a result, which is then transmitted to an endpoint.

Secure models for IoT devices

Consistent distributed edge models via controlled dropout model training are described. According to some embodiments, a machine learning model is trained using multiple sensor data streams. During the training process, ones of the sensor data streams are dropped to cause the model to be generated to be robust in that it can tolerate missing input data from sensor data sources yet still maintain high accuracy. The model can be deployed to multiple sensor devices within an environment. The sensor devices generate sensor data and exchange a variety of types of data to ultimately result in a distributed, consistent model result being generated that remains accurate despite communication faults that may occur between ones of the sensor devices.

Consistent distributed edge models via controlled dropout model training

A deployment service at a remote provider network receives topology data for a local network and generates data filters for edge devices of the local network based on the topology data. The deployment service then sends the data filters to a hub device connected to the local network. The hub device deploys the data filters to respective edge devices of the local network. The data filters may be configured to discard a sufficient portion of collected data to prevent routers from being overloaded by network traffic. The data filters may also be configured to discard a sufficient portion of collected data to prevent the edge devices from consuming too much power in order to preserve energy cost or battery life.

Growing community of inventors

Bellevue, WA, United States of America

Aran Khanna