# US 2021/0176146 A1 · Negotiating machine learning model input features based on cost in constrained networks

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0176146 A1
- Document ID: US20210176146A1
- Status: Published application
- Kind: A1
- Application number: 16/709,235
- Filed: 2019-12-10
- Published: 2021-06-10
- Earliest claimed: 2019-12-10
- Assignees: Cisco Technology, Inc.
- Applicants: Cisco Technology, Inc.
- Inventors: Jean-Philippe Vasseur; Pascal Thubert; Eric Levy-Abegnoli; Patrick Wetterwald
- Web page: https://patentgrep.com/patent/US20210176146A1

## CPC

- H04L 41/145: {involving simulating, designing, planning or modelling of a network}
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 5/04: Inference or reasoning models
- G06N 20/00: Machine learning
- H04L 41/0826: for reduction of network costs (H04L41/0833 takes precedence)
- H04L 41/0853: by actively collecting configuration information or by backing up configuration information
- H04L 43/062: related to network traffic
- G06N 3/04: Architecture, e.g. interconnection topology
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- H04L 41/0631: using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
- H04L 41/147: for predicting network behaviour
- H04L 41/149: for prediction of maintenance
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

In one embodiment, a service receives a feature availability report indicative of which telemetry variables are available at a device in a network and resource costs associated with data features that the device could compute from the telemetry variables. The service selects at least a subset of the data features for input to a machine learning model, based on their associated resource costs and on their respective impacts on one or more performance metrics for the machine learning model. The service trains the machine learning model to evaluate the selected data features. The service sends the trained machine learning model to the device. The device computes the selected data features from the telemetry variables available at the device and uses the computed data features as input to the machine learning model.

## Claims

### Claim 1

A method comprising:
receiving, at a service, a feature availability report indicative of which telemetry variables are available at a device in a network and resource costs associated with data features that the device could compute from the telemetry variables;
selecting, by the service, at least a subset of the data features for input to a machine learning model, based on their associated resource costs and on their respective impacts on one or more performance metrics for the machine learning model;
training, by the service, the machine learning model to evaluate the selected data features; and
sending, by the service, the trained machine learning model to the device, wherein the device computes the selected data features from the telemetry variables available at the device and uses the computed data features as input to the machine learning model.

### Claim 2 (depends on 1)

The method as in claim 1, wherein the one or more performance metrics for the machine learning model are indicative of a recall or precision of the machine learning model.

### Claim 3 (depends on 2)

The method as in claim 2, wherein the service selects the data features for input to the machine learning model to achieve at least a minimum recall or precision.

### Claim 4 (depends on 1)

The method as in claim 1, wherein the feature availability report is further indicative of available resources on the device comprising one or more of: available memory, computing power, or battery of the device, and wherein the service selects the data features for input to the machine learning model based further on the available resources on the device.

### Claim 5 (depends on 1)

The method as in claim 1, further comprising:
determining, by the service, the impacts of the data features on one or more performance metrics for the machine learning model by evaluating how injecting noise into each of the data features affects the one or more performance metrics for the machine learning model.

### Claim 6 (depends on 1)

The method as in claim 1, wherein the resource costs associated with a particular data feature are indicative of resources consumed by the device collecting the telemetry variables from which that data feature is computed.

### Claim 7 (depends on 1)

The method as in claim 1, wherein the machine learning model is configured to make a health inference about the network from the selected data features.

### Claim 8 (depends on 1)

The method as in claim 1, further comprising:
receiving, at the service, feature availability reports from the device over time; and
dynamically adjusting which of the data features are selected for input to the machine learning model, based on the received feature availability reports.

### Claim 9

An apparatus, comprising:
one or more network interfaces to communicate with a network;
a processor coupled to the network interfaces and configured to execute one or more processes; and
a memory configured to store a process executable by the processor, the process when executed configured to:
receive a feature availability report indicative of which telemetry variables are available at a device in a network and resource costs associated with data features that the device could compute from the telemetry variables;
select at least a subset of the data features for input to a machine learning model, based on their associated resource costs and on their respective impacts on one or more performance metrics for the machine learning model;
train the machine learning model to evaluate the selected data features; and
send the trained machine learning model to the device, wherein the device computes the selected data features from the telemetry variables available at the device and uses the computed data features as input to the machine learning model.

### Claim 10 (depends on 9)

The apparatus as in claim 9, wherein the one or more performance metrics for the machine learning model are indicative of a recall or precision of the machine learning model.

### Claim 11 (depends on 10)

The apparatus as in claim 10, wherein the apparatus selects the data features for input to the machine learning model to achieve at least a minimum recall or precision.

### Claim 12 (depends on 9)

The apparatus as in claim 9, wherein the feature availability report is further indicative of available resources on the device comprising one or more of: available memory, computing power, or battery of the device, and wherein the apparatus selects the data features for input to the machine learning model based further on the available resources on the device.

### Claim 13 (depends on 9)

The apparatus as in claim 9, wherein the process when executed is further configured to:
determine the impacts of the data features on one or more performance metrics for the machine learning model by evaluating how injecting noise into each of the data features affects the one or more performance metrics for the machine learning model.

### Claim 14 (depends on 9)

The apparatus as in claim 9, wherein the resource costs associated with a particular data feature are indicative of resources consumed by the device collecting the telemetry variables from which that data feature is computed.

### Claim 15 (depends on 9)

The apparatus as in claim 9, wherein the machine learning model is configured to make a health inference about the network from the selected data features.

### Claim 16 (depends on 9)

The apparatus as in claim 9, wherein the process when executed is further configured to:
receive feature availability reports from the device over time; and
dynamically adjust which of the data features are selected for input to the machine learning model, based on the received feature availability reports.

### Claim 17

A tangible, non-transitory, computer-readable medium storing program instructions that cause a service to execute a process comprising:
receiving, at the service, a feature availability report indicative of which telemetry variables are available at a device in a network and resource costs associated with data features that the device could compute from the telemetry variables;
selecting, by the service, at least a subset of the data features for input to a machine learning model, based on their associated resource costs and on their respective impacts on one or more performance metrics for the machine learning model;
training, by the service, the machine learning model to evaluate the selected data features; and
sending, by the service, the trained machine learning model to the device, wherein the device computes the selected data features from the telemetry variables available at the device and uses the computed data features as input to the machine learning model.

### Claim 18 (depends on 17)

The computer-readable medium as in claim 17, wherein the one or more performance metrics for the machine learning model are indicative of a recall or precision of the machine learning model.

### Claim 19 (depends on 18)

The computer-readable medium as in claim 18, wherein the service selects the data features for input to the machine learning model to achieve at least a minimum recall or precision.

### Claim 20 (depends on 17)

The computer-readable medium as in claim 17, wherein the process further comprises:
determining, by the service, the impacts of the data features on one or more performance metrics for the machine learning model by evaluating how injecting noise into each of the data features affects the one or more performance metrics for the machine learning model.
