Add pole balancing/reinforcement learning tutorial

doc/tutorials/cart_pole_reinforcement_learning/iaf_psc_exp_neuron.nestml

@@ -0,0 +1,58 @@
+# iaf_psc_exp - Leaky integrate-and-fire neuron model
+# ###################################################
+#
+# Description
+# +++++++++++
+#
+# ...
+#
+#
+# References
+# ++++++++++
+#
+# ...
+#
+# See also
+# ++++++++
+#
+model iaf_psc_exp_neuron:
+
+    state:
+        V_m mV = E_l    # Membrane potential
+        g_e real = 0.
+
+    equations:
+        g_e' = -g_e / tau_g
+        V_m' = (g_e * (E_e - V_m) + E_l - V_m + I_e + I_stim) / tau_m
+
+    parameters:
+        tau_m ms = 10 ms          # Membrane time constant
+        tau_g ms = 5 ms
+        E_e mV = 0 mV
+        E_l mV = -74 mV           # Resting potential
+        V_th mV = -54 mV          # Spike threshold potential
+        V_reset mV = -60 mV
+
+        # constant external input current
+        I_e real = 0
+
+        #scaling factor for incoming spikes
+        s real = 1000
+
+    input:
+        spikes_in_port <- spike
+        I_stim real <- continuous
+
+    output:
+        spike
+
+    update:
+        integrate_odes()
+
+    onReceive(spikes_in_port):
+        g_e += spikes_in_port * s
+
+    onCondition(V_m >= V_th):
+        # threshold crossing
+        V_m = V_reset
+        emit_spike()

doc/tutorials/cart_pole_reinforcement_learning/neuromodulated_stdp_synapse.nestml

@@ -0,0 +1,53 @@
+# stdp - Synapse model for spike-timing dependent plasticity
+# ##########################################################
+#
+# ...
+#
+model neuromodulated_stdp_synapse:
+    state:
+        w real = 1    # Synaptic weight
+        wtr real = 0
+        pre_trace real = 0.
+        post_trace real = 0.
+
+    parameters:
+        n real = 0.    # neuromodulator concentration between 0 and 1
+        d ms = 1 ms    # Synaptic transmission delay
+        beta real = 0.01  # learning rate
+        tau_tr_pre ms = 20 ms
+        tau_tr_post ms = 20 ms
+        pre_trace_increment real = 1.
+        post_trace_increment real = 1.
+        wtr_max real = 0.1
+        wtr_min real = 0
+
+    equations:
+        pre_trace' = -pre_trace / tau_tr_pre
+        post_trace' = -post_trace / tau_tr_post
+
+    input:
+        pre_spikes <- spike
+        post_spikes <- spike
+
+    output:
+        spike(weight real, delay ms)
+
+    onReceive(post_spikes):
+        post_trace += -1.05E-7 #  XXX FIXME!!!! should be ``+= post_trace_increment``
+
+        wtr += pre_trace
+        wtr = max(wtr_min, wtr)
+        wtr = min(wtr_max, wtr)
+
+    onReceive(pre_spikes):
+        pre_trace += pre_trace_increment
+
+        wtr += post_trace
+        wtr = max(wtr_min, wtr)
+        wtr = min(wtr_max, wtr)
+
+        # deliver spike to postsynaptic partner
+        emit_spike(w, d)
+
+    update:
+        integrate_odes()