I am currently working on my master's thesis on the topic of applying this type of knowledge to build a reinforcement learning system with spiking neural networks. The role of dopamine is crucial in learning.
By combining spike-timing-dependent plasticity (STDP) with the reward (R-STDP), it is possible to address the spatial eligibility trace (which synapse caused the reward?). However, the design of the reward/utility function is one core issue, where this paper probably advances the field. I am currently looking into ways to balance the reward. Without balancing each trial causes more long term depression than long term potentiation, so that after a while the end the network dies. A different reward function results in the opposite.
Another issue is the distal reward problem: Which event correspondents to which activity?
I would be very glad to discuss related questions or exchange ideas with other experts or newcomers to this growing field. My mail is in my profile.
This paper also got me thinking if this mechanism might explain why we enjoy listening to music. Music follows and breaks rules, which our brain continuously tries to predict and most of the time it succeeds.
By combining spike-timing-dependent plasticity (STDP) with the reward (R-STDP), it is possible to address the spatial eligibility trace (which synapse caused the reward?). However, the design of the reward/utility function is one core issue, where this paper probably advances the field. I am currently looking into ways to balance the reward. Without balancing each trial causes more long term depression than long term potentiation, so that after a while the end the network dies. A different reward function results in the opposite. Another issue is the distal reward problem: Which event correspondents to which activity? I would be very glad to discuss related questions or exchange ideas with other experts or newcomers to this growing field. My mail is in my profile.
This paper also got me thinking if this mechanism might explain why we enjoy listening to music. Music follows and breaks rules, which our brain continuously tries to predict and most of the time it succeeds.