@DLBarack @PessoaBrain The main issue with RNNs in the context of neurophysiology is fading memory, whereas neuronal assemblies can implement persistent processes. This paper attempts to demonstrate how one can use RNNs as substrates for such processes usi
RT @drmichaellevin: After Luis F. Seoane's excellent talk at SFI, am discovering the work of Wolfgang Maass (https://t.co/62GRdKdTIn) - res…
@DLBarack @gualtieropicc @DrYohanJohn @NeuroYogacara @PessoaBrain It is even possible to formalize this mathematically: https://t.co/S1MOEYL886
RT @mraginsky: @IrisVanRooij And this one is even more interesting: https://t.co/S1MOEYL886
@IrisVanRooij And this one is even more interesting: https://t.co/S1MOEYL886
@pfau @drmichaellevin More in line with what Michael is asking, you can apply feedback linearization to an RNN to turn it into a controllable linear system, and then realize your favorite PID controller there: https://t.co/S1MOEYL886
@rtk254 @TaliaRinger Or this: https://t.co/S1MOEYL886
RT @mraginsky: @gershbrain Here is an example of a paper that employs ideas from EE (ODEs, nonlinear control, feedback linearization) and f…
RT @mraginsky: @gershbrain Here is an example of a paper that employs ideas from EE (ODEs, nonlinear control, feedback linearization) and f…
@gershbrain Here is an example of a paper that employs ideas from EE (ODEs, nonlinear control, feedback linearization) and from CS (finite automata, Turing machines) in the context of computational neuroscience: https://t.co/BcfmQJntcC.
@ostojic_srdjan It’s been awhile but I was thinking of research along the lines of https://t.co/fqO08cQSIY
RT @drmichaellevin: After Luis F. Seoane's excellent talk at SFI, am discovering the work of Wolfgang Maass (https://t.co/62GRdKdTIn) - res…
RT @drmichaellevin: After Luis F. Seoane's excellent talk at SFI, am discovering the work of Wolfgang Maass (https://t.co/62GRdKdTIn) - res…
RT @mraginsky: @GaryMarcus @ylecun @tdietterich @andrey_kurenkov @AvilaGarcez Also, feedback in neural nets can be used to simulate symboli…
RT @mraginsky: @GaryMarcus @ylecun @tdietterich @andrey_kurenkov @AvilaGarcez Also, feedback in neural nets can be used to simulate symboli…
@GaryMarcus @ylecun @tdietterich @andrey_kurenkov @AvilaGarcez Also, feedback in neural nets can be used to simulate symbolic computation: https://t.co/BcfmQJntcC 2/2
RT @poipoi2: https://t.co/uMIODizP5g ”... which cannot only simulate any Turing machine, but also any conceivable model for analog computin…
RT @poipoi2: https://t.co/uMIODizP5g ”... which cannot only simulate any Turing machine, but also any conceivable model for analog computin…
https://t.co/uMIODizP5g ”... which cannot only simulate any Turing machine, but also any conceivable model for analog computing with bounded dynamic responses. ” Maass達的にはRNNはチューリング完全どころではないよ，という感じなんだよな
RT @drmichaellevin: After Luis F. Seoane's excellent talk at SFI, am discovering the work of Wolfgang Maass (https://t.co/62GRdKdTIn) - res…
After Luis F. Seoane's excellent talk at SFI, am discovering the work of Wolfgang Maass (https://t.co/62GRdKdTIn) - reservoir computing, liquid state machines, and more. Super-interesting approaches. https://t.co/ErJ2y17Gbx https://t.co/3WzSNKzfRl https:/