Scalable Multi-Task Learning through Spiking Neural Networks with Adaptive Task-Switching Policy for Intelligent Autonomous Agents

Researchers have developed SwitchMT, a novel methodology using Spiking Neural Networks with adaptive task-switching for multi-task learning in autonomous agents. The approach addresses task interference issues and demonstrates competitive performance in multiple Atari games while maintaining low power consumption and network complexity.

• →SwitchMT uses adaptive task-switching instead of fixed intervals to improve multi-task learning performance and scalability.
• →The methodology employs Deep Spiking Q-Networks with active dendrites and task-specific context signals to create specialized sub-networks.
• →Results show competitive scores across multiple Atari games including Pong, Breakout, and Enduro compared to state-of-the-art methods.
• →The approach addresses task interference without increasing network complexity, enabling more efficient autonomous agents.
• →Spiking Neural Networks provide energy-efficient operations through spike-driven data processing for resource-constrained environments.