Differences

This shows you the differences between two versions of the page.

--- examples:sim_background [2014/01/20 10:43] – created zenke
+++ examples:sim_background [2014/03/20 09:04] (current) – adds external link zenke
@@ Line 1: / Line 1: @@
 ====== sim_background ======
-This simulation was used in Zenke et al. 2013 to study the stability of the asynchronous irregular background state in a balanced network in the presence of [[manual:TripletSTDPConnection|TripletSTDP]] over extended periods of time (up to 24h network time; see A in Figure below).
+This simulation is a bit more sophisticated because it comes with many adjustable settings. Therefore the code might not be as straight forward to read as this is the case in the more simple examples. It was used in [[http://www.ploscompbiol.org/article/info%3Adoi%2F10.1371%2Fjournal.pcbi.1003330|Zenke et al. 2013]] to study the stability of the asynchronous irregular background state in a balanced network in the presence of [[manual:TripletSTDPConnection|TripletSTDP]] over extended periods of time (up to 24h network time; see A in Figure below).
 {{:examples:zenke_et_al.png?300|}}
@@ Line 13: / Line 13: @@
 </code>
-Here the parameter ''tau'' is the time constant behind the homeostatic rate estimate, the role of which is center of the study in Zenke et al. 2013. The program will first simulate the non-plastic network for a period of 3*tau seconds before plasticity is enabled. Since the simulation of this plastic network is relatively heavy it offers itself for [[manual:parallel_execution|parallel simulation]].
+Here the parameter ''tau'' is the time constant behind the homeostatic rate estimate, the role of which is center of the study in Zenke et al. 2013. The program will first simulate the non-plastic network for a period of 3*tau seconds before plasticity is enabled. Since the simulation of this plastic network is relatively heavy it offers itself for [[manual:parallel_execution|parallel simulation]]. Suppose you change the parameter ''tau'' to 50 seconds und run the script again the simulation will destabilize after some 10min (cf. Figure above).
 ===== The important bits =====
+Similar to other network simulations such as [[sim_isp_orig]] we start by defining two [[manual:NeuronGroup|neuron groups]]. The ''sim_background'' simulation supports a selection of different neuron models via command line parameters which is why this section looks a little more sophisticated than examples you might have encountered before. The simulation file further allows to specify the conductance ratio between AMPA and NMDA conductance via the function calls ''set_ampa_nmda_ratio(ratio)''
+<code c++>
+    NeuronGroup * neurons_e;
+    if ( adapt ) {
+        neurons_e = new AIFGroup(ne);
+        ((AIFGroup*)neurons_e)->set_ampa_nmda_ratio(ampa_nmda_ratio);
+        ((AIFGroup*)neurons_e)->dg_adapt1=1.0;
+    } else {
+        if ( loadbalance )
+            neurons_e = new IFGroup(ne,1.2,(1.2*ne+ni));
+        else
+            neurons_e = new IFGroup(ne);
+        ((IFGroup*)neurons_e)->set_ampa_nmda_ratio(ampa_nmda_ratio);
+    }
+    IFGroup * neurons_i = new IFGroup(ni);
+    // initialize membranes
+    neurons_i->set_tau_mem(10e-3);
+    neurons_e->random_mem(-60e-3,10e-3);
+    neurons_i->random_mem(-60e-3,10e-3);
+    ((IFGroup*)neurons_i)->set_ampa_nmda_ratio(ampa_nmda_ratio);
+</code>
+The non-plastic connections between the different neuron groups are set up in this section
+<code c++>
+msg = "Setting up I connections ...";
+    logger->msg(msg,PROGRESS,true);
+    SparseConnection * con_ie = new SparseConnection(neurons_i,neurons_e,
+            w_ie,sparseness,GABA);
+    SparseConnection * con_ii = new SparseConnection(neurons_i,neurons_i,
+            w_ii,sparseness,GABA);
+    msg =  "Setting up E connections ...";
+    logger->msg(msg,PROGRESS,true);
+    SparseConnection * con_ei;
+    if ( !ei_plastic ) {
+        con_ei = new SparseConnection(neurons_e,neurons_i,
+                w_ei, sparseness,GLUT);
+    } else {
+        if (infilename.empty()) {
+            con_ei = new TripletConnection(neurons_e,neurons_i,
+                w_ei, sparseness,
+                tau_hom, eta, kappa, wmax, GLUT);
+        } else {
+            string str;
+            str = infilename;
+            str += ".ei.wmat";
+            stringstream oss;
+            oss << "Loading weight matrix from " << str;
+            logger->msg(oss.str(),PROGRESS,true);
+            con_ei = new TripletConnection(neurons_e,neurons_i,
+                str.c_str(),
+                tau_hom, eta, kappa, wmax, GLUT);
+        }
+    }
+</code>
+Finally triplet STDP is setup here. The code again looks a bit more complicated since the simulation supports the use of different flavours of [[manual:TripletSTDPConnection]], the optional addition of Gaussian noise on the initial weights the loading of a predefined connectivity matrix via [[manual:wmat]] file.
+<code c++>
+TripletConnection * con_ee;
+    if (infilename.empty()) {
+        if (decay)
+            con_ee = new TripletDecayConnection(neurons_e,neurons_e,
+                w_ee, sparseness,
+                tau_hom, eta, tau_decay,
+                kappa, wdecay, wmax, GLUT);
+        else
+            con_ee = new TripletConnection(neurons_e,neurons_e,
+                w_ee, sparseness,
+                tau_hom, eta, kappa, wmax, GLUT);
+        // con_ee->set_min_weight(wmin);
+        if ( noisyweights )
+            con_ee->random_data(w_ee,w_ee/4);
+        for ( int i = 0 ; i < n_strengthen ; ++i ) {
+            con_ee->set_data(i,i*(wmax/n_strengthen));
+        }
+    }
+    else {
+        string str;
+        str = infilename;
+        str += ".wmat";
+        stringstream oss;
+        oss << "Loading weight matrix from " << str;
+        logger->msg(oss.str(),PROGRESS,true);
+        if (decay)
+            con_ee = new TripletDecayConnection(neurons_e,neurons_e,
+                str.c_str(),tau_hom,eta,tau_decay,kappa,wdecay,wmax,GLUT);
+        else
+            con_ee = new TripletConnection(neurons_e,neurons_e,
+                str.c_str(),tau_hom,eta,kappa,wmax,GLUT);
+        sprintf(strbuf, "%s.e.nstate", infilename.c_str());
+        neurons_e->load_from_file(strbuf);
+        sprintf(strbuf, "%s.i.nstate", infilename.c_str());
+        neurons_i->load_from_file(strbuf);
+        // primetime = 0;
+    }
+</code>