Time-driven synapse

[beatriiz-barros]

Hi! I'm trying to implement a time-driven synapse, where weights are updated based on a calcium variable post_Ca.

module_name, neuron_model, synapse_model = NESTCodeGeneratorUtils.generate_code_for(nestml_neuron_model="calcium_neuron.nestml",
                                                                                    nestml_synapse_model="time_driven_synapse.nestml",
                                                                                    post_ports=[["post_Ca", "Ca"]],
                                                                                    codegen_opts={"weight_variable": {"time_driven_synapse": "w"},
                                                                                                  "delay_variable": {"time_driven_synapse": "d"},
                                                                                                  "continuous_state_buffering_method": "continuous_time_buffer"})

However, I'm struggling to access that variable inside the synapse model. I was unable to do it directly, so I tried to do it via variable c.

model time_driven_synapse:
    state:
        w real = 0.5

    equations:
        inline c real = post_Ca

    parameters:
        gamma_p real = 8.0
        gamma_d real = 5.0
        theta_p real = 1.3
        theta_d real = 1.0
        tau_w ms = 20.0 ms

        d ms = 1.0 ms

    input:
        pre_spikes <- spike
        post_Ca real <- continuous

    output:
        spike(weight real, delay ms)

    onReceive(pre_spikes):
        emit_spike(w, d)

    update:
        integrate_odes()
        dw real = (- w * (1.0 - w) * (0.5 - w) + gamma_p * (1.0 - w) * heaviside(c, theta_p) - gamma_d * w * heaviside(c, theta_p)) / tau_w
        w += dw * timestep()
        w = max(0.0, min(1.0, w))

    function heaviside(a real, b real) real:
        if a >= b:
            return 1.0
        else:
            return 0.0

Still, I'm getting errors: NESTBuilder_stderr.txt

[pnbabu]

Hi @beatriiz-barros, thanks for writing in! Could you tell us what is the error you get when you try to directly access post_Ca in the synapse?

[beatriiz-barros]

Hi @pnbabu, thank you for your reply!

This is what I get:

"---------------------------------------------------------------------------
CalledProcessError Traceback (most recent call last)
File ~/anaconda3/envs/nest-env/lib/python3.12/site-packages/pynestml/codegeneration/nest_builder.py:181, in NESTBuilder.build(self)
180 current_stage = "build"
--> 181 subprocess.check_call(make_all_cmd, stderr=stderr, stdout=stdout, shell=shell,
182 cwd=str(os.path.join(target_path)))
183 current_stage = "install"

File ~/anaconda3/envs/nest-env/lib/python3.12/subprocess.py:413, in check_call(*popenargs, **kwargs)
412 cmd = popenargs[0]
--> 413 raise CalledProcessError(retcode, cmd)
414 return 0

CalledProcessError: Command '['make', '-j4', 'all']' returned non-zero exit status 2.

During handling of the above exception, another exception occurred:

GeneratedCodeBuildException Traceback (most recent call last)
Cell In[2], line 1
----> 1 module_name, neuron_model, synapse_model = NESTCodeGeneratorUtils.generate_code_for(nestml_neuron_model="calcium_neuron.nestml",
2 nestml_synapse_model="time_driven_synapse.nestml",
3 post_ports=[["post_Ca", "Ca"]],
4 codegen_opts={"weight_variable": {"time_driven_synapse": "w"},
5 "delay_variable": {"time_driven_synapse": "d"},
6 "continuous_state_buffering_method": "continuous_time_buffer"})

File ~/anaconda3/envs/nest-env/lib/python3.12/site-packages/pynestml/codegeneration/nest_code_generator_utils.py:148, in NESTCodeGeneratorUtils.generate_code_for(cls, nestml_neuron_model, nestml_synapse_model, module_name, target_path, post_ports, mod_ports, codegen_opts, logging_level)
145 uniq_id = str(uuid.uuid4().hex)
146 module_name = "nestml_" + uniq_id + "_module"
--> 148 generate_nest_target(input_path=input_fns,
149 install_path=install_path,
150 logging_level=logging_level,
151 module_name=module_name,
152 target_path=target_path,
153 suffix="_nestml",
154 codegen_opts=_codegen_opts)
156 if nestml_synapse_model:
157 return module_name, mangled_neuron_name, mangled_synapse_name

File ~/anaconda3/envs/nest-env/lib/python3.12/site-packages/pynestml/frontend/pynestml_frontend.py:301, in generate_nest_target(input_path, target_path, install_path, logging_level, module_name, store_log, suffix, dev, codegen_opts)
274 def generate_nest_target(input_path: Union[str, Sequence[str]], target_path: Optional[str] = None,
275 install_path: Optional[str] = None, logging_level="ERROR",
276 module_name=None, store_log: bool = False, suffix: str = "",
277 dev: bool = False, codegen_opts: Optional[Mapping[str, Any]] = None):
278 r"""Generate and build code for NEST Simulator.
279
280 Parameters
(...) 299 A dictionary containing additional options for the target code generator.
300 """
--> 301 generate_target(input_path, target_platform="NEST", target_path=target_path, logging_level=logging_level,
302 module_name=module_name, store_log=store_log, suffix=suffix, install_path=install_path,
303 dev=dev, codegen_opts=codegen_opts)

File ~/anaconda3/envs/nest-env/lib/python3.12/site-packages/pynestml/frontend/pynestml_frontend.py:225, in generate_target(input_path, target_platform, target_path, install_path, logging_level, module_name, store_log, suffix, dev, codegen_opts)
191 r"""Generate and build code for the given target platform.
192
193 Parameters
(...) 219 Flag indicating whether errors occurred during processing. False if processing was successful; True if errors occurred in any of the models.
220 """
222 configure_front_end(input_path, target_platform, target_path, install_path, logging_level,
223 module_name, store_log, suffix, dev, codegen_opts)
--> 225 return process()

File ~/anaconda3/envs/nest-env/lib/python3.12/site-packages/pynestml/frontend/pynestml_frontend.py:559, in process()
557 # perform build
558 if _builder is not None:
--> 559 _builder.build()
561 if FrontendConfiguration.store_log:
562 store_log_to_file()

File ~/anaconda3/envs/nest-env/lib/python3.12/site-packages/pynestml/codegeneration/nest_builder.py:188, in NESTBuilder.build(self)
184 subprocess.check_call(make_install_cmd, stderr=stderr, stdout=stdout, shell=shell,
185 cwd=str(os.path.join(target_path)))
187 except subprocess.CalledProcessError as e:
--> 188 raise GeneratedCodeBuildException(stages_exception[current_stage])
190 finally:
191 if self._options["redirect"]:

GeneratedCodeBuildException: Error occurred during 'make all'! More detailed error messages can be found in /home/beatr/hetsyn/plasticity_tests/target/NESTBuilder_stderr.txt."

And the stderr is here: NESTBuilder_stderr.txt.

[pnbabu]

@beatriiz-barros The stderr file isn't complete. Could you attach it again? It would be also helpful if you could share the .nestml files.

[beatriiz-barros]

@pnbabu sorry, I didn't realize. Here are the files again:

NESTBuilder_stderr.txt
calcium_neuron.txt
time_driven_synapse.txt

[pnbabu]

@beatriiz-barros thank you for sharing the files! Looking at the synapse model, the weights are updated with timestep(), does it mean that you require the weights to be updated at every time step? If so, please note that in NEST, the updates in the synapse model is not event-driven, meaning the statements in the update block are executed only on incoming spikes. More information here:https://nestml.readthedocs.io/en/latest/running/running_nest.html#event-based-updating-of-
synapses

Is the synapse model based on any existing literature? If so, a link to the paper would also be helpful.

[beatriiz-barros]

@pnbabu thank you for taking a look at the files! I would like to implement a synapse model where changes in efficacy depend on the postsynaptic calcium concentration, which may change at each timestep, as in Graupner and Brunel, 2012. So, indeed, I would like to have updates on every timestep. Do you think it would be possible to do it with nestml?

But still, why wouldn't I be able to get the value of post_Ca input, even if in an event-driven way?