# US 2021/0342732 A1 · Free-form integration of machine learning model primitives

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

- Number: US 2021/0342732 A1
- Document ID: US20210342732A1
- Status: Published application
- Kind: A1
- Application number: 16/861,671
- Filed: 2020-04-29
- Published: 2021-11-04
- Earliest claimed: 2020-04-29
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Lior Horesh; Giacomo Nannicini; Oktay Gunluk; Sanjeeb Dash; Parikshit Ram; Alexander Gray
- Web page: https://patentgrep.com/patent/US20210342732A1

## CPC

- G06N 20/00: Machine learning
- G06F 7/575: Basic arithmetic logic units, i.e. devices selectable to perform either addition, subtraction or one of several logical operations, using, at least partially, the same circuitry
- G06F 17/00: Digital computing or data processing equipment or methods, specially adapted for specific functions (information retrieval, database structures or file system structures therefor G06F16/00)
- G06F 17/10: Complex mathematical operations {(function generation by table look-up G06F1/03; evaluation of elementary functions by calculation G06F7/544)}
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 5/046: Forward inferencing; Production systems
- G06N 5/02: Knowledge representation; Symbolic representation
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A processor may include a set of primitive operators, receive a set of data-driven operators, at least one of the set of data-driven operators including a machine learning model, and receive an input-output data pair set. Based on a grammar specifying rules for linking the set of primitive operators and the set of data-driven operators, the processor may search among the set of primitive operators and the set of data-driven operators to find a symbolic model that fits the input-output data set.

## Claims

### Claim 1

A computer-implemented method comprising:
receiving a set of primitive operators;
receiving a set of data-driven operators, at least one of the set of data-driven operators including a machine learning model;
receiving an input-output data pair set; and
based on a grammar specifying rules for linking the set of primitive operators and the set of data-driven operators, searching among the set of primitive operators and the set of data-driven operators to find a symbolic model that fits the input-output data set.

### Claim 2 (depends on 1)

The method of claim 1, further including:
defining the grammar specifying the rules for linking the set of primitive operators and the set of data-driven operators.

### Claim 3 (depends on 1)

The method of claim 1, wherein the set of primitive operators include mathematical operators, logical operators and functions.

### Claim 4 (depends on 1)

The method of claim 1, further including training a data-driven operator of the data-driven operators

### Claim 5 (depends on 1)

The method of claim 1, wherein the searching is performed by a constrained optimization search with at least one predefined objective.

### Claim 6 (depends on 5)

The method of claim 5, wherein the predefined objective includes a predefined accuracy objective.

### Claim 7 (depends on 5)

The method of claim 5, wherein the predefined objective includes a predefined expression simplicity objective.

### Claim 8 (depends on 5)

The method of claim 5, wherein the predefined objective includes a fidelity measure.

### Claim 9 (depends on 5)

The method of claim 5, wherein the predefined objective includes an application specific constraint.

### Claim 10

A system comprising:
a hardware processor; and
a memory device coupled with the hardware processor;
the hardware processor configured to at least:
receive a set of primitive operators;
receive a set of data-driven operators, at least one of the set of data-driven operators including a machine learning model;
receive an input-output data pair set;
based on a grammar specifying rules for linking the set of primitive operators and the set of data-driven operators, search among the set of primitive operators and the set of data-driven operators to find a symbolic model that fits the input-output data set.

### Claim 11 (depends on 10)

The system of claim 10, wherein the set of primitive operators include mathematical operators, logical operators, and functions.

### Claim 12 (depends on 10)

The system of claim 10, wherein the hardware processor is configured to train a data-driven operator of the data-driven operators.

### Claim 13 (depends on 10)

The system of claim 10, wherein the hardware processor is configured to perform searching by a constrained optimization search with at least one predefined objective.

### Claim 14 (depends on 13)

The system of claim 13, wherein the predefined objective includes a predefined accuracy objective.

### Claim 15 (depends on 13)

The system of claim 13, wherein the predefined objective includes a predefined expression simplicity objective.

### Claim 16 (depends on 13)

The system of claim 13, wherein the predefined objective includes a fidelity measure.

### Claim 17 (depends on 13)

The system of claim 13, wherein the predefined objective includes an application specific constraint.

### Claim 18

A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a device to cause the device to:
receive a set of primitive operators;
receive a set of data-driven operators, at least one of the set of data-driven operators including a machine learning model;
receive an input-output data pair set; and
based on a grammar specifying rules for linking the set of primitive operators and the set of data-driven operators, search among the set of primitive operators and the set of data-driven operators to find a symbolic model that fits the input-output data set.

### Claim 19 (depends on 18)

The computer program product of claim 18, wherein the set of primitive operators includes mathematical operators, logical operators and functions.

### Claim 20 (depends on 18)

The computer program product of claim 18, wherein the device is caused to train a data-driven operator of the data-driven operators.
