DYNAMIC AND INCREMENTAL EXPLORATION STRATEGY IN FUSION ADAPTIVE RESONANCE THEORY FOR ONLINE REINFORCEMENT LEARNING

Abstract: One of the fundamental challenges in reinforcement learning is to setup a proper balance between exploration and exploitation to obtain the maximum cummulative reward in the long run. Most protocols for exploration bound the overall values to a convergent level of performance. If new knowledge is inserted or the environment is suddenly changed, the issue becomes more intricate as the exploration must compromise the pre-existing knowledge. This paper presents a type of multi-channel adaptive resonance theory (ART) neural network model called fusion ART which serves as a fuzzy approximator for reinforcement learning with inherent features that can regulate the exploration strategy. This intrinsic regulation is driven by the condition of the knowledge learnt so far by the agent. The model offers a stable but incremental reinforcement learning that can involve prior rules as bootstrap knowledge for guiding the agent to select the right action. Experiments in obstacle avoidance and navigation tasks demonstrate that in the configuration of learning wherein the agent learns from scratch, the inherent exploration model in fusion ART model is comparable to the basic E-greedy policy. On the other hand, the model is demonstrated to deal with prior knowledge and strike a balance between exploration and exploitation.

Keywords: Reinforcement Learning, Exploration Strategy, Adaptive Resonance Theory

Author: Budhitama Subagdja

Journal Code: jptkomputergg160009

Artikel Terkait :

Jp Teknik Komputer gg 2016

• An Improved Artificial Bee Colony Algorithm for Staged Search
• Multi-Criteria in Discriminant Analysis to Find the Dominant Features
• A Novel Multifunction Digital Chip Design Based on CMOS Technology
• An Improved Adaptive Niche Differential Evolution Algorithm
• Power Quality Analysis of Integration Photovoltaic Generator to Three Phase Grid under Variable Solar Irradiance Level
• A Combined User-order and Chunk-order Algorithm to Minimize the Average BER for Chunk Allocation in SCFDMA Systems
• An Optimized Model for MapReduce Based on Hadoop
• Hierarchical i* Modeling in Requirement Engineering
• The Optimal High Performance Computing Infrastructure for Solving High Complexity Problem
• Transformer Fault Diagnosis Method Based on Dynamic Weighted Combination Model
• Hybrid Hierarchical Collision Detection Based on Data Reuse
• Brightness and Contrast Modification in Ultrasonography Images Using Edge Detection Results
• Recognition of Fission Signals Based on Wavelet Analysis and Neural Network
• Application of Nonlinear Dynamical Methods for Arc Welding Quality Monitoring
• Chaos-Enhanced Cuckoo Search for Economic Dispatch with Valve Point Effects
• GPU CUDA Accelerated Image Inpainting using Fourth Order PDE Equation
• Comparative Analysis of Spatial Decision Tree Algorithms for Burned Area of Peatland in Rokan Hilir Riau
• Action Recognition of Human’s Lower Limbs Based on a Human Joint
• A Soft Error Study on Tri-gate Based FinFET and Junctionless-FinFET 6T SRAM Cell - A Comparison
• Design of AC Charging Interface and Status Acquisition Circuit for Electric Vehicles
• Towards Smooth and High-Quality Bitrate Adaptation for HTTP Adaptive Streaming
• Features Deletion on Multiple Objects Recognition
• Research on Batch Scheduling in Cloud Computing
• Classification of Motorcyclists not Wear Helmet on Digital Image with Backpropagation Neural Network
• Internet Protocol Based Satellite On-Board System