Deep Reasoning

Deep Reasoning combines deep learning with reasoning for solving complex tasks. It is a critical step toward progress in achieving Artificial General Intelligence.

Deep Reasoning is in early states of research and development. The diagram below illustrates a conceptual architecture:

Attention Processes

Attention mechanisms let a Machine Learning model relate tokens, such as words, to each other regardless of their distance between one another in a group, such as a document or group of documents.

Causal Embedding

Causal Embedding is the process of mapping events and conditions to vectors of real numbers.

Ensemble Learning

Ensemble Learning uses multiple Machine Learning Models to obtain better predictive performance than could be obtained from any single Model.

Input Data

Includes data for:

Known Embedding Inputs

Includes:

  • Phrase and Document Data

  • Lexical Dictionary Data

  • Known Event Sequences

Learned Event Sequences

Memory

Includes:

Modeling System

This is an automated system for performing model training and prediction processing.

Model Training and Prediction Processing

Includes model types such as:

Monitoring and Processes

Monitoring and Processes can be both automated and manual and include the revision and improvement of the Deep Reasoning Modeling System:

  • Architecture

  • Algorithms

  • Accuracy

  • Bias Minimization

  • Input Data

  • Output Data

Output Data

Output data is returned to applications using the Deep Reasoning modeling system.

Predictions

Includes prediction:

  • Data

  • Events

  • Phrases

  • Documents

Word Embedding

Word Embedding is the process of mapping words or phrases to vectors of real numbers.

References