Depends on the target: computational journal rather than neuroscience journal. ## Connection with the future experiments? Earl is collecting data when stimulus is moving left and right directions (perceptually). monkey needs to track continuously moving stimulus, detecting contrast (from left to right or vice versa). Prediction based on human study: if movement is predictable, then the himisphere that is receicing the target will anticipate it to do handshake operation to make sure the transfer to place accuractely (based on EEG in human). But in human study there's not enough resolution, so using monkeys. > Earl: for neuroscience it's interesting cuz it speaks to data. there's also computational inetest to us. Top down: somewhere driving the behavior/learning. ## main contribution synaptic plastiticity explains the phenomenon hebbian long lasting plasticity Akshay thinks we can try to model attractor dynamics to see whether we can explain the same thing. Traditional model are sustaining attractor dynamics. Earl has models that has gaps that still holds memory (due to short term plasticity). That's hypothetical mechnaism of the brain, but people haven't used it; people mainly use attractor dynamics; Earl believes some short term plasticity. Earl wants to start a debate of attractor dynamics vs short term plasticity. He thinks if we can explain convergence due to plasticity, that would be intersting. Long term hebbian is not a part of the debate. > Try recurrence with short term plasticity. > Short term plasticity means more complexity. ## other work (asked by Tommy) Earl with Emily Brown think the way brain working is. The unit that represent content of though is happening at a very fine scale of the brain. Imagine the reoresentation is grain of sands. Checkboard pattern of sandboxes has low andh igh patches of exictability. highs ones get gamma, low ones get beta. info is redundant, so brain does computaiton by checkerboard patterns with diffferent rulings, so compute by changing patterns. brain shifts low frequency oscillations to go to sleep. Take a simple task, high excitability patch stores the pattern. Info in both patches, very redundant, --- Tommy: beta gamma can be an effect of no spiking by synaptic weights? Earl: yea the eclectrc field influences spiking and vice versa. Tommy: block the spike are there waves Earl: spike is a big contributer of the waves. the interplay of field and spiking. cannot get one without the other. spiking and synapses are very important. big idea we are working on. we don't have killer evidence but we have indirect evidence. We have super density neural recording - doing experiments on these. Akshay: top-down control mechanism (yes). allow you to select the area for computation. since you have stored information in many areas; are you selecting patches by beta/gamma activity? Earl: control and content are seperate. top down control and content are represented in the same level with synaptic connections. > Tommy: yea you need model --- Akshay: we can characterize capacity. Earl: could be intersting for working memory capacity. --- ## next steps Hebbian plasciticy is not necessaily the only way do controls in different models --- ##