def check_units():
"""Create a binary, leaky and leaky onset unit to check whether they behave as expected
Turn the binary unit on and off and plot the firing rates"""
units = {}
units["Binary"] = Unit("binary")
units["Binary"].switch_activation_binary_units()
units["Leaky"] = Unit("leaky")
units["Dopaminergic"] = Unit("dopaminergic")
units["Leaky Onset"] = Unit("leaky onset", tau=500, tau_i=500)
units["Leaky"].add_connections([[units["Binary"], 1]])
units["Dopaminergic"].add_connections([[units["Binary"], 1]])
units["Leaky Onset"].add_connections([[units["Binary"], 1]])
t = np.arange(0, 10000, 1)
for i_t in range(1, len(t)):
if i_t == len(t) / 2: # Switch food off again
units["Binary"].switch_activation_binary_units()
for i_u, unit in enumerate(units.values()):
unit.activity(1)
plt.figure()
subplots = np.arange(len(units) * 3).reshape(len(units), 3) + 1
titles = ["Potential", "Firing rate", "Trace"]
for i_u, (name, unit) in enumerate(units.items()):
activity = np.array(unit.activity_history)
for i, title in enumerate(titles):
plt.subplot(len(units), 3, subplots[i_u, i])
plt.plot(activity[:, i])
plt.title(name, fontsize=7)
plt.ylabel(title, fontsize=5)
plt.yticks(fontsize=5)
plt.xticks(fontsize=5)
plt.subplots_adjust(hspace=1)
plt.show()
def build_test_model():
units = {}
units["Binary"] = Unit("binary")
units["Leaky"] = Unit("leaky")
units["Dopaminergic"] = Unit("dopaminergic")
units["Leaky Onset"] = Unit("leaky onset", tau=500, tau_i=500)
units["Leaky"].add_connections([[units["Binary"], 1]])
units["Dopaminergic"].add_connections([[units["Binary"], 1]])
units["Leaky Onset"].add_connections([[units["Binary"], 1]])
# Add the names to the class objects
for name, unit in units.items():
unit.name = name
return units
def test_instrumental_training():
# Define the units needed
units = {}
units["Lever"] = Unit("binary")
units["Food"] = Unit("binary")
units["CS"] = Unit("leaky onset", tau=500, tau_i=500)
units["US"] = Unit("leaky onset", tau=500, tau_i=500)
units["LH"] = Unit("leaky onset", tau=100, tau_i=500)
units["VTA"] = Unit("dopaminergic", dopa_in=0.8, dopa_de=4)
# Connect them
units["CS"].add_connections([[units["Lever"], 5], [units["US"], 0],
[units["VTA"], 1]])
units["US"].add_connections([[units["Food"], 5], [units["CS"], 0],
[units["VTA"],
1]]) # Why deosn't it work that way!
units["LH"].add_connections([[units["Food"], 10], [units["US"], 5]])
units["VTA"].add_connections([[units["LH"], 20]])
# Run the model
t = np.arange(0, 1000000, 1)
lever_on_switches = np.linspace(0, len(t), 10, dtype=int)
food_on_switches = lever_on_switches + 5000
for i_t in range(len(t)):
if i_t in lever_on_switches: # Switch food off again
units["Lever"].switch_activation_binary_units()
elif i_t in lever_on_switches + 10000:
units["Lever"].switch_activation_binary_units()
if i_t in food_on_switches: # Switch food off again
units["Food"].switch_activation_binary_units()
elif i_t in food_on_switches + 10000:
units["Food"].switch_activation_binary_units()
for i_u, unit in enumerate(units.values()):
unit.activity(1)
# Plot the behaviour
plt.figure()
subplots = np.arange(len(units) * 3).reshape(len(units), 3) + 1
titles = ["Potential", "Firing rate", "Trace"]
for i_u, (name, unit) in enumerate(units.items()):
activity = np.array(unit.activity_history)
for i, title in enumerate(titles):
plt.subplot(len(units), 3, subplots[i_u, i])
plt.plot(activity[:, i])
plt.title(name, fontsize=7)
plt.ylabel(title, fontsize=5)
plt.yticks(fontsize=5)
plt.xticks(fontsize=5)
plt.subplots_adjust(hspace=1)
# Plot th weights from CS(Lever) to US(Food) (should increase)
plt.figure()
weights = units["US"].connections[1].weight_history
plt.plot(weights)
plt.title("Weights from CS to US")
plt.show()
print("debug")
def build_BLA_model():
# Define the units needed
units = {}
units["Lever"] = Unit("binary")
units["Food"] = Unit("binary")
units["CS"] = Unit("leaky onset", tau=500, tau_i=500)
units["US"] = Unit("leaky onset", tau=500, tau_i=500)
units["LH"] = Unit("leaky onset", tau=100, tau_i=500)
units["VTA"] = Unit("dopaminergic", dopa_in=0.8, dopa_de=4)
# Connect them
units["CS"].add_connections([[units["Lever"], 5],
[units["US"], 0, "plastic"],
[units["VTA"], 1]])
units["US"].add_connections([[units["Food"],
5], [units["CS"], 0, "plastic"],
[units["VTA"], 1]])
units["LH"].add_connections([[units["Food"], 10], [units["US"], 5]])
units["VTA"].add_connections([[units["LH"], 20]])
# Add the names to the class objects
for name, unit in units.items():
unit.name = name
return units
def build_model():
# Initialize the units of the model
units = {}
units["Cue"] = Unit("binary")
# Amygdala
units["BLA_1"] = Unit("leaky")
# Goal loop
units["NAc_1"] = Unit("leaky")
units["NAc_2"] = Unit("leaky")
units["STNv_1"] = Unit("leaky")
units["STNv_2"] = Unit("leaky")
units["SNpr_1"] = Unit("leaky")
units["SNpr_2"] = Unit("leaky")
units["DM_1"] = Unit("leaky", noise_coeff=15)
units["DM_2"] = Unit("leaky", noise_coeff=15)
units["PL_1"] = Unit("leaky", tau=1000, sigma=2, thres=0)
units["PL_2"] = Unit("leaky", tau=1000, sigma=2, thres=0)
# Action loop
units["DLS_1"] = Unit("leaky")
units["DLS_2"] = Unit("leaky")
units["STNdl_1"] = Unit("leaky")
units["STNdl_2"] = Unit("leaky")
units["GPi_1"] = Unit("leaky")
units["GPi_2"] = Unit("leaky")
units["MGV_1"] = Unit("leaky", noise_coeff=15)
units["MGV_2"] = Unit("leaky", noise_coeff=15)
units["MC_1"] = Unit("leaky", tau=1000, sigma=2, thres=0)
units["MC_2"] = Unit("leaky", tau=1000, sigma=2, thres=0)
# Connect the units - fixed at maximum connection weight
# Goal loop
units["BLA_1"].add_connections([[units["Cue"], 2]])
units["NAc_1"].add_connections([[units["BLA_1"], 2], [units["PL_1"], 1]])
units["NAc_2"].add_connections([[units["PL_2"], 1]])
units["STNv_1"].add_connections([[units["PL_1"], 1]])
units["STNv_2"].add_connections([[units["PL_2"], 1]])
units["SNpr_1"].add_connections([[units["NAc_1"],
-3], [units["STNv_1"], 2],
[units["STNv_2"], 2]])
units["SNpr_2"].add_connections([[units["NAc_2"],
-3], [units["STNv_1"], 2],
[units["STNv_2"], 2]])
units["DM_1"].add_connections([[units["SNpr_1"], -1.5],
[units["DM_2"], -0.8], [units["DM_1"],
0.8]])
units["DM_2"].add_connections([[units["SNpr_2"], -1.5],
[units["DM_1"], -0.8], [units["DM_2"],
0.8]])
units["PL_1"].add_connections([[units["DM_1"], 1]])
units["PL_2"].add_connections([[units["DM_2"], 1]])
units["PL_1"].add_connections([[units["MC_1"], 1]])
units["PL_2"].add_connections([[units["MC_2"], 1]])
# Action loop
units["DLS_1"].add_connections([[units["Cue"], 0], [units["MC_1"], 1]])
units["DLS_2"].add_connections([[units["MC_2"], 1]])
units["STNdl_1"].add_connections([[units["MC_1"], 1]])
units["STNdl_2"].add_connections([[units["MC_2"], 1]])
units["GPi_1"].add_connections([[units["DLS_1"],
-3], [units["STNdl_1"], 2],
[units["STNdl_2"], 2]])
units["GPi_2"].add_connections([[units["DLS_2"],
-3], [units["STNdl_1"], 2],
[units["STNdl_2"], 2]])
units["MGV_1"].add_connections([[units["GPi_1"], -1.5],
[units["MGV_1"], 0.8],
[units["MGV_2"], -0.8]])
units["MGV_2"].add_connections([[units["GPi_2"], -1.5],
[units["MGV_1"], -0.8],
[units["MGV_2"], 0.8]])
units["MC_1"].add_connections([[units["MGV_1"], 1]])
units["MC_2"].add_connections([[units["MGV_2"], 1]])
units["MC_1"].add_connections([[units["PL_1"], 0.2]])
units["MC_2"].add_connections([[units["PL_2"], 0.2]])
# Add the names to the class objects
for name, unit in units.items():
unit.name = name
return units
def build_goal_loop():
# Initialize the units of the model (start with the goal loop)
units = {}
units["Cue"] = Unit("binary")
units["Amygdala"] = Unit("binary")
units["VTA"] = Unit("dopaminergic", dopa_in=0.8, dopa_de=4)
units["NAc_1"] = Unit("leaky")
units["NAc_2"] = Unit("leaky")
units["STNv_1"] = Unit("leaky")
units["STNv_2"] = Unit("leaky")
units["SNpr_1"] = Unit("leaky")
units["SNpr_2"] = Unit("leaky")
units["DM_1"] = Unit("leaky", noise_coeff=6)
units["DM_2"] = Unit("leaky", noise_coeff=6)
units["PL_1"] = Unit("leaky", tau=2000, sigma=20, thres=0.8)
units["PL_2"] = Unit("leaky", tau=2000, sigma=20, thres=0.8)
# Connect the units - fixed at maximum connection weight
units["VTA"].add_connections([[units["Amygdala"], 2]])
# What connection weight between teh cue and the NAc? In the MM the maximum connection weight was 2 for each and
# each US was connected to each NAc --> Hence a connection from the cue to both NAc with max weight 4?
units["NAc_1"].add_connections([[units["Cue"], 2], [units["PL_1"], 1]])
units["NAc_2"].add_connections([[units["Cue"], 0.5], [units["VTA"], 1],
[units["PL_2"], 1]])
units["STNv_1"].add_connections([[units["PL_1"], 1]])
units["STNv_2"].add_connections([[units["PL_2"], 1]])
units["SNpr_1"].add_connections([[units["NAc_1"],
-3], [units["STNv_1"], 2],
[units["STNv_2"], 2]])
units["SNpr_2"].add_connections([[units["NAc_2"],
-3], [units["STNv_1"], 2],
[units["STNv_2"], 2]])
# What is the weight of the connection of a DM unit on itself? Chose default: 1
units["DM_1"].add_connections([[units["SNpr_1"], -1.5],
[units["DM_2"], -0.8], [units["DM_1"], 1]])
units["DM_2"].add_connections([[units["SNpr_2"], -1.5],
[units["DM_1"], -0.8], [units["DM_2"], 1]])
units["PL_1"].add_connections([[units["DM_1"], 1]])
units["PL_2"].add_connections([[units["DM_2"], 1]])
# Add the names to the class objects
for name, unit in units.items():
unit.name = name
return units
#######################################################################################################
# Get the weight of the input (dependent on dopaminergic input units)
def get_input_weight(self):
# Check if the unit is connected to a dopaminergic unit
dopa_input_unit = self.get_dopa_input_unit()
if dopa_input_unit:
input_weight = dopa_input_unit.dopa_in + dopa_input_unit.dopa_de * dopa_input_unit.firing_rate
else:
input_weight = 1
return input_weight
#######################################################################################################
# Return the dopaminergic input unit if there is one, else none
def get_dopa_input_unit(self):
dopa_input_unit = [
connection.input_unit for connection in self.connections
if connection.input_unit.type == "dopaminergic"
]
if dopa_input_unit:
return dopa_input_unit[0]
else:
return None
# Update the weights - only for units that receive dopaminergic input
dopa_input_unit = self.get_dopa_input_unit()
if dopa_input_unit:
for connection in self.connections:
# Update only plastic connections
if connection.type == "plastic":
# Leaky onset units (BLA learning)
if self.type == "leaky onset":
weight_change = self.eta_bla * pos_sat(dopa_input_unit.firing_rate - self.thres_da_bla) * \
pos_sat(self.trace_change) * neg_sat(connection.input_unit.trace_change) * \
(self.max_w_bla - connection.weight)
# Leaky units (Striatal learning)
elif self.type == "leaky":
weight_change = self.eta_str * pos_sat(dopa_input_unit.firing_rate - self.thres_da_str) * \
pos_sat(self.firing_rate - self.thres_str) *\
pos_sat(connection.input_unit.firing_rate - self.thres_inp_str)
connection.weight = connection.weight + weight_change * dt
# Save the connection weight in an array
connection.weight_history.append(connection.weight)
def build_fixed_action_loop():
# Initialize the units of the model, leave out the amygdala for now
units = {}
units["Cue"] = Unit("binary")
units["Amgydala"] = Unit("binary")
units["DLS_1"] = Unit("leaky")
units["DLS_2"] = Unit("leaky")
units["STNdl_1"] = Unit("leaky")
units["STNdl_2"] = Unit("leaky")
units["GPi_1"] = Unit("leaky")
units["GPi_2"] = Unit("leaky")
units["MGV_1"] = Unit("leaky", noise_coeff=6)
units["MGV_2"] = Unit("leaky", noise_coeff=6)
units["MC_1"] = Unit("leaky", tau=2000, sigma=20, thres=0.8)
units["MC_2"] = Unit("leaky", tau=2000, sigma=20, thres=0.8)
# Connect the units
units["DLS_1"].add_connections([[units["Amgydala"], -1], [units["Cue"], 1],
[units["MC_1"], 1]])
units["DLS_2"].add_connections([[units["Amgydala"], 1], [units["MC_2"],
1]])
units["STNdl_1"].add_connections([[units["MC_1"], 1.6]])
units["STNdl_2"].add_connections([[units["MC_2"], 1.6]])
units["GPi_1"].add_connections([[units["DLS_1"],
-3], [units["STNdl_1"], 2],
[units["STNdl_2"], 2]])
units["GPi_2"].add_connections([[units["DLS_2"],
-3], [units["STNdl_1"], 2],
[units["STNdl_2"], 2]])
units["MGV_1"].add_connections([[units["GPi_1"],
-1.5], [units["MGV_1"], 1],
[units["MGV_2"], -0.8]])
units["MGV_2"].add_connections([[units["GPi_2"], -1.5],
[units["MGV_1"], -0.8],
[units["MGV_2"], 1]])
units["MC_1"].add_connections([[units["MGV_1"], 1]])
units["MC_2"].add_connections([[units["MGV_2"], 1]])
# Add the names to the class objects
for name, unit in units.items():
unit.name = name
return units
def build_fixed_model():
# Initialize the units of the model (start with the goal loop)
units = {}
units["Lever"] = Unit("binary")
units["Chain"] = Unit("binary")
units["FoodA"] = Unit("binary")
units["FoodB"] = Unit("binary")
units["SatA"] = Unit("binary")
units["SatB"] = Unit("binary")
units["CSa"] = Unit("leaky onset", tau=500, tau_i=500)
units["CSb"] = Unit("leaky onset", tau=500, tau_i=500)
units["USa"] = Unit("leaky onset", tau=500, tau_i=500)
units["USb"] = Unit("leaky onset", tau=500, tau_i=500)
units["LH"] = Unit("leaky onset", tau=100, tau_i=500)
units["VTA"] = Unit("dopaminergic", dopa_in=0.8, dopa_de=1.5)
units["NAc_1"] = Unit("leaky")
units["NAc_2"] = Unit("leaky")
units["STNv_1"] = Unit("leaky")
units["STNv_2"] = Unit("leaky")
units["SNpr_1"] = Unit("leaky")
units["SNpr_2"] = Unit("leaky")
units["DM_1"] = Unit("leaky", noise_coeff=6)
units["DM_2"] = Unit("leaky", noise_coeff=6)
units["PL_1"] = Unit("leaky", tau=2000, sigma=20, thres=0.8)
units["PL_2"] = Unit("leaky", tau=2000, sigma=20, thres=0.8)
# Connect the units
# BLA units connected to themselves?
units["CSa"].add_connections([[units["Lever"], 5], [units["CSb"], 2],
[units["USa"], 2], [units["USb"], 2],
[units["VTA"], 1]])
units["CSb"].add_connections([[units["Chain"], 5], [units["CSa"], 2],
[units["USa"], 2], [units["USb"], 2],
[units["VTA"], 1]])
units["USa"].add_connections([[units["FoodA"], 5], [units["FoodB"], 5],
[units["SatA"], -10], [units["SatB"], -10],
[units["CSa"], 2], [units["CSb"], 2],
[units["USb"], 2], [units["VTA"], 1]])
units["USb"].add_connections([[units["FoodA"], 5], [units["FoodB"], 5],
[units["SatA"], -10], [units["SatB"], -10],
[units["CSa"], 2], [units["CSb"], 2],
[units["USa"], 2], [units["VTA"], 1]])
units["LH"].add_connections([[units["FoodA"], 10], [units["FoodB"], 10],
[units["USa"], 5], [units["USb"], 5]])
units["VTA"].add_connections([[units["LH"], 20]])
units["NAc_1"].add_connections([[units["VTA"], 1], [units["USa"], 2],
[units["USb"], 2], [units["PL_1"], 1]])
units["NAc_2"].add_connections([[units["VTA"], 1], [units["USa"], 2],
[units["USb"], 2], [units["PL_2"], 1]])
units["STNv_1"].add_connections([[units["PL_1"], 1.6]])
units["STNv_2"].add_connections([[units["PL_2"], 1.6]])
units["SNpr_1"].add_connections([[units["NAc_1"],
-3], [units["STNv_1"], 2],
[units["STNv_2"], 2]])
units["SNpr_2"].add_connections([[units["NAc_2"],
-3], [units["STNv_1"], 2],
[units["STNv_2"], 2]])
units["DM_1"].add_connections([[units["SNpr_1"], -1.5], [units["DM_1"], 1],
[units["DM_2"], -0.8]])
units["DM_2"].add_connections([[units["SNpr_2"], -1.5],
[units["DM_1"], -0.8], [units["DM_2"], 1]])
units["PL_1"].add_connections([[units["DM_1"], 1]])
units["PL_2"].add_connections([[units["DM_2"], 1]])
# Add the names to the class objects
for name, unit in units.items():
unit.name = name
return units