Connectome function

Functional model testing.

Tasks

class connectome.function.modeltest.Task

Abstract base class for tasks.

All tasks should derive from this class.

abstract __call__(model) → dict
Parameters

model (Node) –

A model instance of ER, EXP, SORN, LL, API, ERFEVER or SYN. Please note the following correspondences between the class names used in this code package and the model acronyms used in the manuscript (class name: acronym):

  • ER: ER-ESN,

  • EXP: EXP-LSM,

  • LL: LAYERED,

  • SYN: SYNFIRE,

  • ERFEVER: FEVER,

  • API: API,

  • SORN: STDP-SORN.

Returns

task_result – Dictionary of task results.

Return type

dict

class connectome.function.task.memory.MemoryTask(simulation_time=20)

Short term memory task. Test the short term memory property of a network.

Accepts network models with feature vectors. E.g., API and FEVER are accepted.

The task representation is obtained as the sum over the neurons’ feature vectors weighted by their activity.

Parameters

simulation_time (float) – Simulation time relative to the neuronal time constant,

class connectome.function.task.texture.TextureTask(nb_epoch=6000, nr_samples=10000, nr_classes=7, length=500, verbose=0, learning_rate=<function default_learning_rate>, slice_last=250)

Texture discrimination task.

Can be run with any network model.

The reported accuracy is the accuracy of each output time step considered as individual test point.

Parameters

  • nb_epoch (int) – Number epochs for training.

  • nr_samples (int) – Total number of samples for the task.

  • nr_classes (int) – Number of texture classes sampled from natural images.

  • length (int) – Length of a texture sequence.

  • verbose (int) – How much information to print during training.

  • learning_rate (callable) – Learning_rate(nr_exc_layers) should return the learning rate for a model with nr_exc_layers excitatory layers.

get_data()
Returns

X, y – Data, one-hot-encoded target.

Return type

tuple of arrays

class connectome.function.task.tuning.TuningTask(return_arrays=False)

Stimulus tuning task.

Checks if the cortical tuning is sharper than the thalamic input.

Accepts network models with feature vectors. E.g. FEVER and API are accepted.

class connectome.function.task.unsynchronized.UnsynchronizedActivityTask

Accepts the SORN model.

Checks if the activity of the network is synchronized or not.

class connectome.function.task.propagation.task.PropagationTask(run_time=100.0 * msecond, spike_detection_pool_fraction: float = 0.5, **dyn_syn_pars)

Checks the propagation of activity along a synfire chain.

Accepts the SYN.

Parameters

  • run_time (brian2.Quantity) – Simulation time.

  • spike_detection_pool_fraction (float in [0, 1]) – Fraction of neurons in a pool which have to spike to detect event as pool activation.

Criteria

class connectome.function.modeltest.Criterion

Abstract Criterion base class. All criteria should implement the interface defined in this ABC.

abstract __call__(value: Union[float, bool]) → bool
Parameters

value (float or bool) – The value to be evaluated.

Returns

passed – Returns true if the criterion was satisfied.

Return type

bool

class connectome.function.criterion.GreaterThan(lower)

Greater than criterion.

True if the test quantity is greater than lower.

Parameters

lower (float) – Quantity has to be strictly greater than lower to yield “true”.

class connectome.function.criterion.LessThan(upper)

Greater than criterion.

True if the test quantity is less then upper.

Parameters

upper (float) – Quantity has to be strictly less then upper to yield true.

class connectome.function.criterion.Range(lower, upper)

Range criterion.

True if the test quantity is within the range [lower, upper].

Parameters

  • lower (float) – Lower boundary of range.

  • upper (float) – Upper boundary of range.

class connectome.function.criterion.TrueCriterion

Return True if the test quantity itself is True.

Test runners

class connectome.function.modeltest.TestSuite(task: connectome.function.modeltest.Task)

Test suite for functional checking.

Parameters

task (Task) – The task for this test suite.

add_criterion(criterion_name: str, criterion: connectome.function.modeltest.Criterion)

Add a test criterion.

Parameters

  • criterion_name (str) – Name of the quantity to be tested. This has to correspond to one of the output keys of the task.

  • criterion (Criterion) – The test criteria.

enqueue_model(model)

Enqueue a model to be tested, but do not execute the tests.

Parameters

model (Model) – Can be run by the model_runner. Usually an instance of one of the network models.

execute_enqueued_models(mapper=None) → connectome.function.modeltest.TestSuiteResultList

Execute the test task on all enqueued models.

Parameters

mapper (map-like object) – Can be distributed. Has to work like a map.

Returns

test_suite_result_list – List of results of the test.

Return type

TestSuiteResultList

mapper

alias of map

test_model(model) → connectome.function.modeltest.TestSuiteResult

Test a single model.

Parameters

model (Model) – Can be run by the model_runner. Usually an instance of one of the network models.

Test results

class connectome.function.modeltest.TestSuiteResultList(results: List[connectome.function.modeltest.TestSuiteResult])

Represents a list of test suite results.

Parameters

results (List[TestSuiteResult]) – A list of test suite results.

append(item: connectome.function.modeltest.TestSuiteResult)
Parameters

item (TestSuiteResult) – Append item to the list.

done()

Check if the tests have executed already. This is meant to work with the Future protocol as in the concurrent.futures module.

Returns

done – True if done, false otherwise.

Return type

bool

property passed

True if all tests were passed.

property results_list

Transforms the possibly passed iterator into a list on first call and returns it.

wait()

Wait for all tasks to be executed and return then.

class connectome.function.modeltest.TestSuiteResult(model: str)

Result of a single test.

Parameters

model (Node) – The evaluated model.

add(evaluation: connectome.function.modeltest.TestResult)

Add a single test result of the evaluation of a single criterion to the test suite result.

Parameters

evaluation (TestResult) – Result of a single test.

property passed

True if all criteria were passed.

class connectome.function.modeltest.TestResult(criterion_name: str, criterion: Optional[connectome.function.modeltest.Criterion], score: Union[float, KeyError], passed: bool)

Result of the evaluation of a single criterion.

Parameters

  • criterion_name (str) – Name of the checked criterion.

  • criterion (Union[Criterion, None]) – The checked criterion.

  • score (float) – The score of the criterion.