Hierarchical Memory System for Generalized Pattern Recognition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current artificial intelligence approaches, such as classic AI and neural networks, fail to replicate human intelligence due to their lack of understanding of the essence of intelligence and simplistic emulation of neural networks, which ignores the complex anatomy of the human brain, limiting their ability to generalize and understand the world.

Innovation Solution

A hierarchical memory system with cortical processing units that process sequences of patterns, using a single cortical algorithm to store and recognize invariant structures, and make predictions based on stored memories, mimicking the human brain's neocortex organization and operation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If neural networks are used to replicate human intelligence, then the ability to learn and generalize is improved, but the system remains primitive and lacks real understanding of the world

Engineering Contradiction:
Improveability to learn and generalizeVSAvoidreal understanding of the world
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The invention divides the brain into separate functional modules: sensory input processing, memory storage, pattern recognition, and prediction generation. Each module handles specific functions, allowing the system to achieve both learning capability and meaningful understanding by processing information through distinct computational stages rather than using a monolithic neural network

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The invention introduces a temporal dimension by processing information as sequences of patterns over time, rather than static inputs. This sequential processing allows the system to capture dynamics and context, enabling it to make predictions about future states and understand causal relationships, thereby achieving more reliable understanding while maintaining adaptability

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Measurement precision

If classic AI algorithms are used to solve specific problems, then problem-solving capability is improved, but the system cannot learn or generalize to novel inputs

Engineering Contradiction:
Improveproblem-solving capabilityVSAvoidability to learn and generalize
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary learning by storing sequences of patterns in memory during the training phase. These stored sequences serve as pre-processed knowledge that can be quickly retrieved and applied to novel problems, allowing the system to both solve specific problems precisely and generalize to new situations without requiring retraining

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The invention implements feedback mechanisms where the system compares predicted outputs with actual outcomes and uses this information to refine its pattern recognition and prediction capabilities. This feedback loop enables continuous improvement of both problem-solving accuracy and generalization ability, resolving the contradiction between precision and adaptability

Inventive Principle:
Principle #23Feedback

3Adaptability or versatility

If a hierarchical memory system with cortical processing units is implemented, then pattern recognition and prediction abilities are improved, but device complexity increases

Engineering Contradiction:
Improvepattern recognition and predictionVSAvoidhierarchical structure
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The invention creates a universal hierarchical processing architecture where the same basic cortical processing units can be applied across different domains (visual, auditory, tactile) by simply changing the input data format. This multi-functional design allows the system to achieve advanced pattern recognition and prediction capabilities while controlling complexity through standardized, reusable modules rather than custom-designed systems for each function

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentUS9530091B2Methods, architecture, and apparatus for implementing machine intelligence and hierarchical memory systems
Publication Date: 2016.12.27 NUMENTA INC
  • US9530091B2 patent drawing
  • US9530091B2 patent drawing
  • US9530091B2 patent drawing

AI summary

Sophisticated memory systems and intelligent machines may be constructed by creating an active memory system with a hierarchical architecture. Specifically, a system may comprise a plurality of individual cortical processing units arranged into a hierarchical structure. Each individual cortical processing unit receives a sequence of patterns as input. Each cortical processing unit processes the received input sequence of patterns using a memory containing previously encountered sequences with structure and outputs another pattern. As several input sequences are processed by a cortical processing unit, it will therefore generate a sequence of patterns on its output. The sequence of patterns on its output may be passed as an input to one or more cortical processing units in next higher layer of the hierarchy. A lowest layer of cortical processing units may receive sensory input from the outside world. The sensory input also comprises a sequence of patterns.