IPAB Workshop - 05/04/2018 Speaker Tianqi Wei Titile A Bio-inspired Reinforcement Learning Rule to Optimise Dynamical Neural Networks for Robot Control Abstract Most approaches for optimisation of neural networks are based on variants of back-propagation. This requires the network to be time invariant and differentiable; neural networks with dynamics are thus generally outside the scope of these methods. Biological neural circuits are highly dynamic yet clearly able to support learning. We propose a reinforcement learning approach inspired by the mechanisms and dynamics of biological synapses. The network weights undergo spontaneous fluctuations, and a reward signal modulates the centre and amplitude of fluctuations to converge to a desired network behaviour. We test the new learning rule on a 2D bipedal walking simulation, using a control system that combines a recurrent neural network, a bio-inspired central pattern generator layer and proportional-integral control, and demonstrate the first successful solution to this benchmark task. Speaker Marija Jegorova Title GANs Application to Robot Control Abstract Current Generative Adversarial Networks (GANs) applications mostly include image or video generation and completion. I will report my recent progress on applying GANs to generating trajectories for a robotic arm (specifically Baxter). Producing a selection of diverse trajectories for throwing to the same set target is potentially applicable to obstacle avoidance. Apr 05 2018 12.45 - 14.00 IPAB Workshop - 05/04/2018 Tianqi Wei, Marija Jegorova and Jane Loveless IF 4.31/4.33
IPAB Workshop - 05/04/2018 Speaker Tianqi Wei Titile A Bio-inspired Reinforcement Learning Rule to Optimise Dynamical Neural Networks for Robot Control Abstract Most approaches for optimisation of neural networks are based on variants of back-propagation. This requires the network to be time invariant and differentiable; neural networks with dynamics are thus generally outside the scope of these methods. Biological neural circuits are highly dynamic yet clearly able to support learning. We propose a reinforcement learning approach inspired by the mechanisms and dynamics of biological synapses. The network weights undergo spontaneous fluctuations, and a reward signal modulates the centre and amplitude of fluctuations to converge to a desired network behaviour. We test the new learning rule on a 2D bipedal walking simulation, using a control system that combines a recurrent neural network, a bio-inspired central pattern generator layer and proportional-integral control, and demonstrate the first successful solution to this benchmark task. Speaker Marija Jegorova Title GANs Application to Robot Control Abstract Current Generative Adversarial Networks (GANs) applications mostly include image or video generation and completion. I will report my recent progress on applying GANs to generating trajectories for a robotic arm (specifically Baxter). Producing a selection of diverse trajectories for throwing to the same set target is potentially applicable to obstacle avoidance. Apr 05 2018 12.45 - 14.00 IPAB Workshop - 05/04/2018 Tianqi Wei, Marija Jegorova and Jane Loveless IF 4.31/4.33
