Talk:Models of tactile perception and development
The review article is concise, informative and provides a good cross-section of the state-of-the-art in models of tactile perception and development. This will make an excellent contribution to the scholarpedia project.
Other than the minor typos listed below I think a cross-reference to the recent Scholarpedia article Pearson and Pipe "whiskered robots" 10(3):6641 in section "Model application" would be appropriate.
Typos: 1. Section, Model application: 2nd to last line typo "3strategies" remove "3". 2. Section, Biological application: "decreasingly shorter" should read as "increasingly shorter"
=
A clear and concise review that summarizes a large body of work from the last years on tactile perception. A few comments below, minor comments were sent directly to the author.
(1) Fig 1 – are the ‘perceptual goals’ really changing based on sensory input, or – the perceptual goals are pre-determined, and the motor strategies/outputs Change with new sensory info? (2) Reinforcement Vs optimal control – I think it will be helpful to say in a sentence or two what is the fundamental difference. “completion of the task” in the case of “tactile perception” may well be “maximizing perception”, in which case, from your description it seems that the difference is that ‘reinforcement learning’ does not take into account the cost. Any other fundamental difference? (3) I think that the description of the human whisker experiment is very nice, but it does not stress enough the point that active sensing strategy was chosen, even though it is definitely not the only possible strategy. I would say that subject could solve the task by positioning their hands over the pole and sensing the angular difference between the hands, however, subject choose an active sensing strategy...just like rodents... (4) The paragraph "tactile navigation/biological application" should be expended a bit to explain the rational underlying the model. why "...motor strategy that balances between exploration motor primitives and retreats" ? Isn't 'retreat' just another 'motor primitives'? why novelty? and why novelty management (balance between novelty and retreat..) ? It seems to pop out of the blue, the way it is currently presented. I think you should explain the rational, make clear the idea that the novelty is the reward (this is a new idea..) and explain how is it related in a fundamental way to active sensing. (5) The two parts of 'tactical navigation' seems somewhat unrelated to each other. I would try to link the two in a closing remark. Is there any common link between the two described models?