Thoughts(Unfinished)
In contrast to the research interests on the main page, here is a more specific elaboration
Let's start with a list of paper topics that I've (carefully) read (mostly theorists). After that I'll explain why I got my point of view from these readings. If I've been biased in some of my views because I didn't realize some of the work existed, please email me to let me know!
List: William Bialek, Larry Abbott(Connectome), Haim Sompolinsky(Manifold Capacity, Cerebellum-like structure, Chaotic NN), Terry Sejnowski(Traveling Wave), Ken Miller(V1), Xiao-Jing Wang(Multi-regional Interaction), Stefano Fusi(Abstract Geometry), Ashok Litwin-Kumar(Connectome), James Fitzgerald(Zebrafish), Christof Koch(Books about conciousness), Dmitry Chklovski, Tatiana Engel, Stephanie Palmer, Brent Doiron, Surya Ganguli, Yonatan Aljadeff, Marcus Benna, Tatyana Sharpee.................................
First let me clarify the general interets of me:
Short-term: Let the data speak. Quantify the neural data, build phenomenological desciption.
Long-term: Build local and global biophysics model based on observed phenomenological model.
About both short-term and long-term interets, I already have some preference about research method and potential mechanism. They are Maximal Entropy Model, Coarse-grained analysis, and Traveling wave.
I have a general map of such theory: in the local circuits, neurons operate in a new computation mechanism. In the global circuits, between the cortical and subcortical regions, some mechanism connect them(traveling wave). While the structure in each minicolumn connectivity motif is important. Then such structure support the conciousness. (Later I will explain the reason)
There is a sentence about how to choose research interests (objects): choose the smallest object that could assess to understands our original goals.
One of the things that particularly disappoints and frustrates me about modern neuroscience (both experimental and computational) is that the vast majority of research is devoted to making the answers to questions more complex, as if we start with a very crude model and keep adding embellishments to it in anticipation that the model will someday be equivalent to a real brain. But with this methodology, we don't know when we've reached the end of the line. (I don't even think such a methodology will ever reach the end.)
List: William Bialek, Larry Abbott(Connectome), Haim Sompolinsky(Manifold Capacity, Cerebellum-like structure, Chaotic NN), Terry Sejnowski(Traveling Wave), Ken Miller(V1), Xiao-Jing Wang(Multi-regional Interaction), Stefano Fusi(Abstract Geometry), Ashok Litwin-Kumar(Connectome), James Fitzgerald(Zebrafish), Christof Koch(Books about conciousness), Dmitry Chklovski, Tatiana Engel, Stephanie Palmer, Brent Doiron, Surya Ganguli, Yonatan Aljadeff, Marcus Benna, Tatyana Sharpee.................................
First let me clarify the general interets of me:
Short-term: Let the data speak. Quantify the neural data, build phenomenological desciption.
Long-term: Build local and global biophysics model based on observed phenomenological model.
About both short-term and long-term interets, I already have some preference about research method and potential mechanism. They are Maximal Entropy Model, Coarse-grained analysis, and Traveling wave.
I have a general map of such theory: in the local circuits, neurons operate in a new computation mechanism. In the global circuits, between the cortical and subcortical regions, some mechanism connect them(traveling wave). While the structure in each minicolumn connectivity motif is important. Then such structure support the conciousness. (Later I will explain the reason)
There is a sentence about how to choose research interests (objects): choose the smallest object that could assess to understands our original goals.
One of the things that particularly disappoints and frustrates me about modern neuroscience (both experimental and computational) is that the vast majority of research is devoted to making the answers to questions more complex, as if we start with a very crude model and keep adding embellishments to it in anticipation that the model will someday be equivalent to a real brain. But with this methodology, we don't know when we've reached the end of the line. (I don't even think such a methodology will ever reach the end.)
