17523553. REINFORCEMENT LEARNING WITH INDUCTIVE LOGIC PROGRAMMING simplified abstract (International Business Machines Corporation)
REINFORCEMENT LEARNING WITH INDUCTIVE LOGIC PROGRAMMING
Organization Name
International Business Machines Corporation
Inventor(s)
Songtao Lu of White Plains NY (US)
REINFORCEMENT LEARNING WITH INDUCTIVE LOGIC PROGRAMMING - A simplified explanation of the abstract
This abstract first appeared for US patent application 17523553 titled 'REINFORCEMENT LEARNING WITH INDUCTIVE LOGIC PROGRAMMING
Simplified Explanation
Methods and systems are disclosed for training a model and automated motion using reinforcement learning. The approach involves learning Markov decision processes in training environments and extracting logic rules from them. These logic rules are then used to train a reward logic neural network (LNN) and a safety LNN. The reward LNN and safety LNN take a state-action pair as input and output a corresponding score for the pair.
- Learning Markov decision processes using reinforcement learning in training environments
- Extracting logic rules from the learned Markov decision processes
- Training a reward logic neural network (LNN) and a safety LNN using the extracted logic rules
- The reward LNN and safety LNN provide scores for state-action pairs
Potential Applications
- Autonomous vehicles: Training models for self-driving cars to make decisions based on logic rules extracted from Markov decision processes.
- Robotics: Teaching robots to perform complex tasks by training them using reinforcement learning and logic rules.
- Gaming: Developing intelligent game agents that can learn and make decisions based on logic rules.
Problems Solved
- Complex decision-making: The technology enables training models to make complex decisions by extracting logic rules from Markov decision processes.
- Safety concerns: The safety LNN helps ensure that the trained models prioritize safety in their actions.
- Efficiency: By using reinforcement learning and logic rules, the training process can be more efficient and effective.
Benefits
- Improved decision-making: The trained models can make better decisions by considering both rewards and safety factors.
- Automation: The technology enables automated motion by training models to perform tasks without human intervention.
- Versatility: The approach can be applied to various domains, such as autonomous vehicles, robotics, and gaming.
Original Abstract Submitted
Methods and systems for training a model and automated motion include learning Markov decision processes using reinforcement learning in respective training environments. Logic rules are extracted from the Markov decision processes. T reward logic neural network (LNN) and a safety LNN are trained using the logic rules extracted from the Markov decision processes. The reward LNN and the safety LNN each take a state-action pair as an input and output a corresponding score for the state-action pair.