Reinforcement learning of targeted movement (Chadderdon et al. 2012)

 Download zip file   Auto-launch 
Help downloading and running models
Accession:144538
"Sensorimotor control has traditionally been considered from a control theory perspective, without relation to neurobiology. In contrast, here we utilized a spiking-neuron model of motor cortex and trained it to perform a simple movement task, which consisted of rotating a single-joint “forearm” to a target. Learning was based on a reinforcement mechanism analogous to that of the dopamine system. This provided a global reward or punishment signal in response to decreasing or increasing distance from hand to target, respectively. Output was partially driven by Poisson motor babbling, creating stochastic movements that could then be shaped by learning. The virtual forearm consisted of a single segment rotated around an elbow joint, controlled by flexor and extensor muscles. ..."
Reference:
1 . Chadderdon GL, Neymotin SA, Kerr CC, Lytton WW (2012) Reinforcement learning of targeted movement in a spiking neuronal model of motor cortex PLoS ONE 2012 7(10):e47251
Model Information (Click on a link to find other models with that property)
Model Type: Realistic Network;
Brain Region(s)/Organism: Neocortex;
Cell Type(s): Neocortex fast spiking (FS) interneuron; Neocortex spiking regular (RS) neuron; Neocortex spiking low threshold (LTS) neuron;
Channel(s):
Gap Junctions:
Receptor(s): GabaA; AMPA; NMDA;
Gene(s):
Transmitter(s): Dopamine; Gaba; Glutamate;
Simulation Environment: NEURON;
Model Concept(s): Simplified Models; Synaptic Plasticity; Long-term Synaptic Plasticity; Reinforcement Learning; Reward-modulated STDP;
Implementer(s): Neymotin, Sam [samn at neurosim.downstate.edu]; Chadderdon, George [gchadder3 at gmail.com];
Search NeuronDB for information about:  GabaA; AMPA; NMDA; Dopamine; Gaba; Glutamate;
/
arm1d
README
drspk.mod *
infot.mod *
intf6_.mod *
intfsw.mod *
misc.mod *
nstim.mod *
stats.mod *
updown.mod *
vecst.mod *
arm.hoc
basestdp.hoc
col.hoc *
colors.hoc *
declist.hoc *
decmat.hoc *
decnqs.hoc *
decvec.hoc *
default.hoc *
drline.hoc *
filtutils.hoc *
geom.hoc
grvec.hoc *
hinton.hoc *
infot.hoc *
init.hoc
intfsw.hoc *
labels.hoc *
local.hoc *
misc.h *
mosinit.hoc
network.hoc
nload.hoc
nqs.hoc *
nqsnet.hoc *
nrnoc.hoc *
params.hoc
run.hoc
samutils.hoc *
sense.hoc *
setup.hoc *
sim.hoc
simctrl.hoc *
stats.hoc *
stim.hoc
syncode.hoc *
units.hoc *
xgetargs.hoc *
                            
// $Id: run.hoc,v 1.92 2012/03/05 21:58:50 samn Exp $

declare("wf1","o[1]","wf2","o[1]","wrec","o[1]")
sprint(tstr,"o[%d]",numcols)
declare("vit",tstr,"nqLFP",tstr,"SPKSZ",25e6*tstop/20e3,"vitdp","o[1]","maxrec",1)

method("local")
cvode.atol(1e-3) 
cvode.condition_order(1) // irrelevant to acells?
declare("sepflds",0) // whether to record separate fields for each layer

// rv2 is part of grvec.hoc, as a stub, so it can be
// customized to do something to vector before graphing it
proc rv2 () { }

//** draw lines between cell subpopulations
proc rasterlines () { local i,j,ct,c localobj o
  {g=graphItem c=CDX}
  for CDX=0,numcols-1 for ct=0,CTYPi-1 if(col[CDX].numc[ct]) {
    if(ct==DP) { 
      j=cedp.o(0).id
      i=cedp.o(cedp.count-1).id
    } else {
      i=col[CDX].ce.o(col[CDX].ixe[ct]).gid
      j=col[CDX].ce.o(col[CDX].ix[ct]).gid
    }
    drline(0,i,tstop,i,g,2,6)
    o=mdl2view(g,0.9,j+col[CDX].numc[ct]/4)
    g.label(0.9,o.x[3],CTYP.o(ct).s)
  }
  {CDX=c g.flush}
}
//** calls rasterlines, making sure g is set first
proc grlines () { local id
  if(numarg()>0) id=$1 else id=0
  {g=Graph[id] rasterlines()}
}

//** a
proc a () { local sh,sv
  if(g==nil)gg()
  if (!isobj(aa,"Graph")) aa=g else g=aa
  if (aa.view_count==0) aa=g
  sv=gnum gnum=ojtnum(g)
  graphItem=g
  sh=0 grv_.super=1 g.erase_all
  grv_.gveraseflag=0  grv_.gvmarkflag=grv_.super=1 // gnum=0
  gv(0,1+sh,2) // gv(1,3+sh,2) gv(2,2+sh,2)
  grv_.gvmarkflag=grv_.super=0
  gnum=sv
  rasterlines()
}

//** b
proc b () { rewt() time() a() }

//** initrr - for doing a rerun - not used right now
proc initrr () { 
  rewt()
  col.intf.global_init()
  NStim[0].global_init()
  vseed_stats(392426)
}

//** setMemb - nothing here
proc setMemb () {}

//** initMisc1
proc initMisc1 () { local i localobj xo
  vseed_stats(392426)
  col.intf.global_init()
  for i=0,numcols-1 col[i].cstim.initrands()
  for ltr(xo,cedp) xo.soma.v = vrefrac_DRSPK //make sure DPD voltages initialized to same level 
}

objref ww // global ww for post processing
{wwht_INTF6=1 wwwid_INTF6=100}
//** wrecon - setup LFP recording, one LFP for each COLUMN
proc wrecon () { local cdx,ii,jj,x,ct,cnt localobj tl,vm
  tl=new List()
  for cdx=0,numcols-1 {
    nqsdel(nqLFP[cdx])
    if(sepflds) nqLFP[cdx]=new NQS("ES","EM","LFP") else {nqLFP[cdx]=new NQS(1) nqLFP[cdx].s[0].s="LFP"}
    {nqLFP[cdx].v.resize(tstop/vdt_INTF6) nqLFP[cdx].pad()}
    for ii=0,nqLFP[cdx].m-1 tl.append(nqLFP[cdx].v[ii])
  }
  {vm=new Vector(CTYPi) vm.x(ES)=0 vm.x(EM)=1}
  col.intf.initwrec(tl)
  for x=0,numcols-1 {
    for case(&ct,ES,EM,&jj) {
      for ii=col[x].ix[ct],col[x].ixe[ct] {
        if(sepflds) {
          col[x].ce.o(ii).wrc(vm.x(jj)+x*3)
          col[x].ce.o(ii).wrc(x*3)
        } else col[x].ce.o(ii).wrc(x)
      }
    }
  }
}

//** wrecoff - turn off LFP recording
proc wrecoff () { local ct
  for ctt(&ct) for ixt(ct) XO.wrec(0)
}

//** finishMisc - called @ end of run()
func finishMisc () { local ii localobj co
  for ltr(XO,printlist) if (isassigned(XO.o)) if (XO.o.fflag) XO.o.fini
  // if(0) panobj.pvplist(ofile,params,100) //dont save printlist for now
  col.intf.global_fini
  for ltr(co,lcol) co.intf.spkstats2(1)
  print "TMAX: ",tmax_INTF6
  return 1
}

//** snapsv() save after printlist items min-max to fixed dt
proc snapsv () { local a,vdt,min,max localobj v1,o
  grv_.bst(3,3)
  vdt=0.2 
  a=allocvecs(v1)
  v1.resize(tstop/vdt)
  for ltr(o,printlist) { 
    if (o.code!=3) continue
    v1.snap(o.vec,o.tvec,vdt)
    o.vec.copy(v1)
    o.pstep=vdt
    o.tvflag=0
  }
  if (numarg()==0) grv_.pvall()
  dealloc(a)
}

proc exeruncall () { for ltr(XO,printlist) if (XO.code==3) XO.tvflag=1 }
proc pvout2 () { snapsv(1) }

//printlist=new List()
if(printlist==nil)printlist=new List()
proc prlclr () { localobj ce,col,intf
  for ltr(XO,printlist) {
    if (isassigned(XO.o)) if (XO.o.fflag) XO.o.recclr
  }
  //  for ltr(XO,ce) XO.wrc(-1)
  for ltr(col,lcol) {
    {ce=col.ce intf=ce.o(0)}
    for ii=0,ce.count-1 ce.o(ii).wrc(-1)
    intf.wwfree(0)
  }
  printlist.remove_all
}

//** prl(recv,recs[,lvextra]) - setup recording in printlist
//$1  = whether to record any cell voltages, default off
//$2  = whether to record spike times, default on
//$o3 = extra cells to record. lv.o(0)=cell,lv.o(1)=param to record,etc.,optional
proc prl () { local a,x,cidx,ii,jj,offst,y,recv,recs,max localobj xo,lvextra,co,ce
  if(numarg()>0) recv=$1 else recv=0
  if(numarg()>1) recs=$2 else recs=1
  if(numarg()>2) lvextra=$o3 else lvextra=nil
  {offst=0 prlclr()}
  for ltr(co,lcol,&ii) {
    {ce=co.ce intf=co.intf}
    if (recs) { sprint(tstr,"%s_SPKS",co.name)
      if (intf.flag("jcn")) { // for use with jitcon()
        printlist.append((vit[ii]=new vitem(tstr,SPKSZ,1)))
        intf.jitrec(vit[ii].vec,vit[ii].tvec)
      } else {
        intf.jitrec() // clear jit recording
        for ltr(xo,ce,&y) {
          if (y==0) vit[ii]=new_printlist_nc(xo, xo.id, tstr) else {
            new_printlist_nc2(vit[ii], xo, xo.id)        }
        }
      }
    }
    if(co.numc[DP]>0) {
      printlist.append(vitdp=new vitem("DPSPKS",100))
      for ltr(xo,cedp,&y) new_printlist_nc(xo.drspk, y+co.ix[DP])
//      for ltr(xo,cedp,&y) new_printlist_nc(xo.drspk, xo.id) // y+co.ix[DP])
    }
    npacsz=20
    if (recv && ce.count>0) {
      for x=0,CTYPi-1 if(co.numc[x]>0 && x!=DP) {
        max=maxrec-1
        for jj=0,max {
          XO=ce.o(co.ix[x]+jj)
          XO.recclr
          new_printlist_ac(XO,"V",  CTYP.o(x).s,XO.id)
          // new_printlist_ac(XO,"VGB",  CTYP.o(x).s,XO.id)
          // new_printlist_ac(XO,"VGA",  CTYP.o(x).s,XO.id)
          // new_printlist_ac(XO,"AHP",  CTYP.o(x).s,XO.id)
          //      printlist.o(printlist.count-1).code=3 // use code 3 for snapping
          // XO=ce.object(ixe[x]-jj-1)
          // XO.recclr
          // new_printlist_ac(XO,"V",  CTYP.object(x).s,XO.id)
        }
      }
    }
    if(lvextra!=nil){
      for(jj=0;jj<lvextra.count;jj+=2){ XO=lvextra.o(jj)
        new_printlist_ac(XO,lvextra.o(jj+1).s,CTYP.object(ctyp(XO.id)).s,XO.id)
      }
    }
  }
  wrecon()
}
 
stopoq_INTF6=1

//** turnoff
proc turnoff () { local cel0,cel1,off
  cel0=$1 cel1=$2
  if (argtype(3)==0) off=$3 else off=0
  ind.indgen(ix[cel0],ixe[cel0],1) vec.indgen(ix[cel1],ixe[cel1],1) 
  intf.turnoff(ind,vec,off)
}

//** turn off intralaminar connections
proc intralamoff () { local ct,i
  for i=0,numcols-1 for col[i].ctt(&ct) if(col[i].numc[ct] && col[i].div[ct][ct]) turnoff(ct,ct)
}

//** turn on intralaminar connections
proc intralamon () { local ct,i
  for i=0,numcols-1 for col[i].ctt(&ct) if(col[i].numc[ct] && col[i].div[ct][ct]) turnoff(ct,ct,1)
}

//* mkrecl(col,ty) - setup recording of spikes from a population of cells
// return List with 2 sub-Lists: 1st with NetCons, 2nd with Vectors storing spikes
obfunc mkrecl () { local ct localobj xo,nc,col,ls,myncl,myspkl,myv
  col=$o1 ct=$2
  ls=new List()
  ls.append(myncl=new List())
  ls.append(myspkl=new List())
  for ltr(xo,col.ce) {
    if(xo.type!=ct) continue
    xo.flag("out",1) // make sure NetCon events enabled from this cell
    myncl.append(nc=new NetCon(xo,nil))
    myspkl.append(myv=new Vector())
    nc.record(myv) // record each cell separately
  }
  return ls
}

//* function calls
prl(1,1)

Loading data, please wait...