Synaptic scaling balances learning in a spiking model of neocortex (Rowan & Neymotin 2013)

 Download zip file   Auto-launch 
Help downloading and running models
Accession:147141
Learning in the brain requires complementary mechanisms: potentiation and activity-dependent homeostatic scaling. We introduce synaptic scaling to a biologically-realistic spiking model of neocortex which can learn changes in oscillatory rhythms using STDP, and show that scaling is necessary to balance both positive and negative changes in input from potentiation and atrophy. We discuss some of the issues that arise when considering synaptic scaling in such a model, and show that scaling regulates activity whilst allowing learning to remain unaltered.
Reference:
1 . Rowan MS, Neymotin SA (2013) Synaptic Scaling Balances Learning in a Spiking Model of Neocortex Adaptive and Natural Computing Algorithms, Tomassini M, Antonioni A, Daolio F, Buesser P, ed. pp.20
Citations  Citation Browser
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 L5/6 pyramidal GLU cell; Neocortex L2/3 pyramidal GLU cell; Neocortex V1 interneuron basket PV GABA cell; Neocortex fast spiking (FS) interneuron; Neocortex spiny stellate cell; Neocortex spiking regular (RS) neuron; Neocortex spiking low threshold (LTS) neuron; Abstract integrate-and-fire adaptive exponential (AdEx) neuron;
Channel(s):
Gap Junctions:
Receptor(s): GabaA; AMPA; NMDA;
Gene(s):
Transmitter(s): Gaba; Glutamate;
Simulation Environment: NEURON; Python;
Model Concept(s): Synaptic Plasticity; Long-term Synaptic Plasticity; Learning; STDP; Homeostasis;
Implementer(s): Lytton, William [bill.lytton at downstate.edu]; Neymotin, Sam [Samuel.Neymotin at nki.rfmh.org]; Rowan, Mark [m.s.rowan at cs.bham.ac.uk];
Search NeuronDB for information about:  Neocortex L5/6 pyramidal GLU cell; Neocortex L2/3 pyramidal GLU cell; Neocortex V1 interneuron basket PV GABA cell; GabaA; AMPA; NMDA; Gaba; Glutamate;
/
stdpscalingpaper
batchscripts
mod
README
alz.hoc
autotune.hoc *
basestdp.hoc *
batch.hoc *
batch2.hoc *
batchcommon
checkirreg.hoc *
clusterrun.sh
col.dot *
col.hoc *
comppowspec.hoc *
condisconcellfig.hoc *
condisconpowfig.hoc *
declist.hoc *
decmat.hoc *
decnqs.hoc *
decvec.hoc *
default.hoc *
drline.hoc *
e2hubsdisconpow.hoc *
e2incconpow.hoc *
filtutils.hoc *
geom.hoc *
graphplug.hoc *
grvec.hoc *
init.hoc *
labels.hoc *
load.hoc *
local.hoc *
makepopspikenq.hoc *
matfftpowplug.hoc *
matpmtmplug.hoc *
matpmtmsubpopplug.hoc *
matspecplug.hoc *
network.hoc *
nload.hoc *
nqpplug.hoc *
nqs.hoc *
nqsnet.hoc *
nrnoc.hoc *
params.hoc
plot.py
plotbatch.sh
plotbatchcluster.sh
powchgtest.hoc *
python.hoc *
pywrap.hoc *
redE2.hoc *
run.hoc
runsim.sh
setup.hoc *
shufmua.hoc *
sim.hoc
simctrl.hoc *
spkts.hoc *
stats.hoc *
stdpscaling.hoc
syncode.hoc *
vsampenplug.hoc *
writedata.hoc
xgetargs.hoc *
                            
// $Id: autotune.hoc,v 1.1 2011/04/17 17:40:22 samn Exp $

//* sim duration
tstop=mytstop=htmax=6000e3

//* load sim

EEGain = EIGain = IEGain = IIGain = 1

autotune = 1 

load_file("nqsnet.hoc")
load_file("network.hoc")
load_file("params.hoc")
load_file("run.hoc")
load_file("nload.hoc")

jrtm_INTF6 = tstop + 1 // less printouts from intf6.mod

//* declares

declare("myncl",new List(),"myspkl",new List(),"myspktyl",new List(),"vice",new Vector(col.allcells))
declare("vtarg",new Vector(col.allcells))
declare("syl","o[2]") // list of sywvs

declare("nqrat","o[1]")
declare("ltab","o[1]","contab","o[1]","wtab","o[2]","mtab","o[1]")
declare("veinlyr",new Vector(CTYPi)) // corresponding E cells for I cells of a layer
veinlyr.x(I2)=E2
veinlyr.x(I2L)=E2
veinlyr.x(I4)=E4
veinlyr.x(I4L)=E4
veinlyr.x(I5)=E5R
veinlyr.x(I5L)=E5R
veinlyr.x(I6)=E6
veinlyr.x(I6L)=E6

declare("witer",0,"updinc",2e3,"uprob",0.1)
declare("EEinc",0.01,"EIinc",0.01,"IEinc",0.01,"IIinc",0.01,"skipI",1)

declare("vrecw",new Vector(),"nqrec","o[1]")

declare("IFctr",1.01,"IRate",4,"ERate",1)

//* lowrefrac - lower refractory period
proc lowrefrac () { local i localobj xo
  // lower refrac to allow more flexible freq. alterations
  for i=0,numcols-1 for ltr(xo,col[i].ce) if(!ice(xo.type))xo.refrac=2.5 else xo.refrac=1.25
}

//* settarg - set target # of spikes (per period)
proc settarg () { local i localobj xo
  for ltr(xo,col.ce,&i) {
    if(ice(xo.type)) vtarg.x(i) = IRate else vtarg.x(i) = ERate
  }
  vtarg.mul(updinc/1e3)
}

//* settunerc - setup recording of spikes used in tuning
proc settunerec () { local i localobj xo,nc
  for i=0,CTYPi-1 myspktyl.append(new Vector())
  for ltr(xo,col.ce,&i) {
    xo.flag("out",1) // make sure NetCon events enabled from this cell
    myncl.append(nc=new NetCon(xo,nil))
    myspkl.append(new Vector())
    nc.record(myspkl.o(i)) // record each cell separately
    vice.x(i)=ice(xo.type)
  }
}

//* mksyl - setup lists of weight vectors
proc mksyl () { local i,dvt localobj vw1,vw2
  for i=0,1 syl[i]=new List()
  for i=0,col.allcells-1 {
    dvt=col.ce.o(i).getdvi()
    vw1=new Vector(dvt)
    vw2=new Vector(dvt)
    col.ce.o(i).getsywv(vw1,vw2)
    syl[0].append(vw1)
    syl[1].append(vw2)
  }
}

//* conn2tab - make lookup tables with connectivity info
obfunc conn2tab () { local i,j,k,id1,id2 localobj ltab,col,nqc,contab,wtab1,wtab2,mtab,vc
  col=$o1 ltab=new List() vc=new Vector(col.allcells)
  for i=0,3 ltab.append(new Matrix(col.allcells,col.allcells))
  {contab=ltab.o(0) wtab1=ltab.o(1) wtab2=ltab.o(2) mtab=ltab.o(3)}
  if(col.connsnq==nil) {
    print "conn2tab ERR: col.connsnq is nil"
    return nil
  }
  nqc=col.connsnq
  nqc.sort("del") // make sure order in NQS corresponds to getdvi order
  for i=0,nqc.v.size-1 {
    id1=nqc.v[0].x(i) // from id1
    id2=nqc.v[1].x(i) // to id2
    contab.x(id1,id2)=1 // is there a connection?
    wtab1.x(id1,id2)=nqc.v[4].x(i) // weight 1
    wtab2.x(id1,id2)=nqc.v[5].x(i) // weight 2
    mtab.x(id1,id2) = vc.x(id1) // index into div vector -- assumes order in connsnq according to div
    vc.x(id1) += 1
  }
  return ltab
}

//* updatewts - update the weights
proc updatewts () { local i,j,md,df,fctr,inc,w0,w1,idx,trg,ety,cidx,a localobj xo,vs,ce
  print "t:", t, ", witer:",witer
  //a=allocvecs(vs)
  ce=col.ce
  for i=0,CTYPi-1 if(col.numc[i]) myspktyl.o(i).resize(0) //setup per-type counts
  for ltr(xo,ce,&i) myspktyl.o(xo.type).append(myspkl.o(i))
  for ltr(xo,ce,&i) {
//    if(rdm.uniform(0,1) < uprob) continue
    if(vice.x(i)) { // presynaptic I cell      
      for j=0,col.allcells-1 if(contab.x(i,j) && myspkl.o(i).size) {
        if(vice.x(j)) { // postsynaptic I cell
          if(skipI) continue
          inc = IIinc
          ety = veinlyr.x(ce.o(j).type)
          trg = MAXxy(vtarg.x(j),IFctr*myspktyl.o(ety).size/col.numc[ety])
        } else { // postsynaptic E cell
          inc = IEinc
          trg = vtarg.x(j)
        }
        df = myspkl.o(j).size - trg
        if(df > 0) fctr = 1 + inc else if(df < 0) {
          fctr = 1 - inc
        }
        idx = mtab.x(i,j)
        w0 = syl[0].o(i).x(idx) * fctr
        if(w0 < 0) syl[0].o(i).x(idx)=wtab[0].x(i,j)=0 else syl[0].o(i).x(idx)=wtab[0].x(i,j)=w0
      }
      xo.setsywv(syl[0].o(i),syl[1].o(i)) // reset weights
    } else { // presynaptic E cell
      for j=0,col.allcells-1 if(contab.x(i,j) && myspkl.o(i).size) {
        if(vice.x(j)) { // postsynaptic I cell
          if(skipI) continue
          inc = EIinc
          ety = veinlyr.x(ce.o(j).type)
          trg = MAXxy(vtarg.x(j),IFctr*myspktyl.o(ety).size/col.numc[ety])
        } else { // postsynaptic E cell
          inc = EEinc
          trg = vtarg.x(j)
        }
        df = myspkl.o(j).size - trg
        if(df > 0) fctr = 1 - inc else  if(df < 0) {
          fctr = 1 + inc
        }
        idx = mtab.x(i,j) // index into div,weight array
        w0 = syl[0].o(i).x(idx) * fctr // updated weight
        if(w0 < 0) syl[0].o(i).x(idx)=wtab[0].x(i,j)=0 else syl[0].o(i).x(idx)=wtab[0].x(i,j)=w0
        syl[1].o(i).x(idx) = wtab[1].x(i,j) = syl[0].o(i).x(idx) * NMAMR // NMDA weight
      }
      xo.setsywv(syl[0].o(i),syl[1].o(i)) // reset weights
    }
  }
  for vtr(&i,vrecw) {
    for j=0,col.allcells-1 if(contab.x(i,j)) {
      idx = mtab.x(i,j)
      nqrec.append(i,j,ce.o(i).type,ce.o(j).type,syl[0].o(i).x(idx),syl[1].o(i).x(idx),witer)
    }
  }
  witer += 1
  for i=0,myspkl.count-1 myspkl.o(i).resize(0) // reset spike counts for cells to 0
  for(i=CTYPi-1;i>=0;i-=1) if(col.numc[i]) {
    j=1e3*myspktyl.o(i).size/(col.numc[i]*updinc)
    print CTYP.o(i).s, " " , j , " avg Hz "
    nqrat.append(witer,i,j)
    myspktyl.o(i).resize(0) // reset spike counts for types to 0
  }
  //dealloc(a)
  cvode.event(t+updinc,"updatewts()") // set next update weights event
}

//* mysend - starts off the update q
proc mysend () { 
  {nqsdel(nqrat) nqrat=new NQS("witer","ty","rate")}
  {nqsdel(nqrec) nqrec=new NQS("id1","id2","ty1","ty2","w0","w1","witer")}
  cvode.event(updinc,"updatewts()") 
}
declare("fith",new FInitializeHandler("mysend()"))

//* calls

lowrefrac()
settarg()
settunerec()
mksyl()

ltab=conn2tab(col)
contab=ltab.o(0)
wtab[0]=ltab.o(1)
wtab[1]=ltab.o(2)
mtab=ltab.o(3)