# US 2014/0156577 A1 · Methods and systems for artificial cognition

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

- Number: US 2014/0156577 A1
- Document ID: US20140156577A1
- Status: Published application
- Kind: A1
- Application number: 14/094,142
- Filed: 2013-12-02
- Published: 2014-06-05
- Earliest claimed: 2012-12-05
- Assignees: Applied Brain Research Inc.
- Applicants: Applied Brain Research Inc
- Inventors: Christopher David Eliasmith; Terrence Charles Stewart; Feng-Xuan Choo; Trevor William Bekolay; Travis Crncich-DeWolf; Yichuan Tang; Daniel Halden Rasmussen
- Web page: https://patentgrep.com/patent/US20140156577A1

## CPC

- G06N 3/092: Reinforcement learning
- G06N 3/042: Knowledge-based neural networks; Logical representations of neural networks
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/049: Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
- G06N 3/0495: Quantised networks; Sparse networks; Compressed networks
- G06N 3/082: modifying the architecture, e.g. adding, deleting or silencing nodes or connections
- G06N 3/09: Supervised learning
- G06N 3/10: Interfaces, programming languages or software development kits, e.g. for simulating neural networks
- G06N 20/00: Machine learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Methods, systems and apparatus that provide for perceptual, cognitive, and motor behaviors in an integrated system implemented using neural architectures. Components of the system communicate using artificial neurons that implement neural networks. The connections between these networks form representations—referred to as semantic pointers—which model the various firing patterns of biological neural network connections. Semantic pointers can be thought of as elements of a neural vector space, and can implement a form of abstraction level filtering or compression, in which high-dimensional structures can be abstracted one or more times thereby reducing the number of dimensions needed to represent a particular structure.

## Claims

### Claim 1

An artificial intelligence system comprising:
at least one interface hierarchy configured to receive an input of a high-dimensional representation and to compress the high-dimensional representation to generate a lower-dimensional representation of the input;
at least one processing module configured to receive the lower-dimensional representation and to generate a further representation;
an action selection controller configured to control communication of the lower-dimensional representation and the further representation between the at least one interface hierarchy and the at least one processing module.

### Claim 2 (depends on 1)

The system of claim 1, wherein an additional interface hierarchy is configured to decompress the further representation to generate an additional high-dimensional representation, the additional high-dimensional representation usable to produce an output.

### Claim 3 (depends on 1)

The system of claim 1, wherein the at least one hierarchy is implemented using a neural network.

### Claim 4 (depends on 3)

The system of claim 3, wherein the neural network is a spiking neural network.

### Claim 5 (depends on 1)

The system of claim 1, wherein the at least one processing module is implemented using a neural network.

### Claim 6 (depends on 5)

The system of claim 5, wherein the neural network is a spiking neural network.

### Claim 7 (depends on 1)

The system of claim 1, wherein the action selection controller is implemented using a neural network.

### Claim 8 (depends on 7)

The system of claim 7, wherein the neural network is a spiking neural network.

### Claim 9 (depends on 1)

The system of claim 1, further comprising at least one control point, wherein the action selection controller controls communication by modulating the at least one control point.

### Claim 10 (depends on 1)

The system of claim 1, wherein the at least one hierarchy comprises one or more modules selected from the list consisting of a visual input module, a working memory module and a motor output module.

### Claim 11 (depends on 1)

The system of claim 1, wherein the at least one processing module comprises one or more modules selected from the list consisting of an information encoding module, a transform calculation module, a reward evaluation module, an information decoding module and a motor processing module.

### Claim 12 (depends on 1)

The system of claim 1, wherein the action selection controller comprises one or more hierarchical processing networks.

### Claim 13 (depends on 12)

The system of claim 12, wherein the one or more hierarchical processing networks have functions analogous to at least one anatomical structure of mammalian basal ganglia.

### Claim 14 (depends on 13)

The system of claim 13, wherein the at least one anatomical structure is selected from the group consisting of striatum, subthalmic nucleus, ventral striatum, globus pallidus externus, globus pallidus internus, substantia nigra pars reticulate, substantia nigra pars compacta, and ventral tegmental area.

### Claim 15

An artificial intelligence system comprising:
at least one perceptual module configured to compress input received by the system to generate an internal representation;
at least one cortical module configured to process the internal representation;
an action selection controller configured to coordinate communication of the internal representation between the at least one perceptual module, the at least one cortical module and the at least one output module.

### Claim 16 (depends on 15)

The system of claim 15, wherein the at least one perceptual module is further configured to decompress the internal representation.

### Claim 17 (depends on 15)

The system of claim 15, wherein the at least one perceptual module is composed of spiking artificial neurons.

### Claim 18 (depends on 15)

The system of claim 15, wherein the at least one perceptual module is composed of non-spiking artificial neurons.

### Claim 19 (depends on 15)

The system of claim 15, wherein the at least one cortical module is composed of spiking artificial neurons.

### Claim 20 (depends on 15)

The system of claim 15, wherein the at least one cortical module is composed of non-spiking artificial neurons.

### Claim 21 (depends on 15)

The system of claim 15, wherein the action selection controller is composed of spiking artificial neurons.

### Claim 22 (depends on 15)

The system of claim 15, wherein the action selection controller is composed of non-spiking artificial neurons.

### Claim 23 (depends on 15)

The system of claim 15, further comprising at least one motor module configured to decompress the transformed internal representation.

### Claim 24 (depends on 23)

The system of claim 23, wherein the at least one motor module is further configured to compress the internal representation.

### Claim 25 (depends on 23)

The system of claim 23, wherein the at least one motor module is composed of spiking artificial neurons.

### Claim 26 (depends on 23)

The system of claim 23, wherein the at least one motor module is composed of non-spiking artificial neurons.

### Claim 27 (depends on 15)

The system of claim 15, wherein the internal representation is a semantic pointer representation.

### Claim 28 (depends on 15)

The system of claim 15, wherein the at least one cortical module processes the internal representation using superposition and a compression operator, wherein the compression operator is selected from the group consisting of: circular convolution, element-wise multiplication, principle component extraction, and learned operators.