Below, you can find a linked version of the changelog available at the PyRates github repository.
This official release is a combination of all the bug fixes and improvements in the pre 1.0 versions.
Minor improvements since 0.17.4:
removed typos in documentation
improved layout of the online documentation
updated documentation to account for latest changes
removed bug where a delayed differential equation model would lead to a KeyError when trying to generate the run function
added sign function to the backend that returns the sign of its input (1 or -1)
improved readthedocs documentation (removed bug with display of math, added new use example for edge templates)
added a safe guard for defining edge templates: An error is raised now when edge template input variables have the same name as their source variable.
minor debugging of the model introduction use examples
adjustments of the template cheat sheet template_specification.rst
debugged issue in base backend, where file names specified by the users that contained a file ending wre not handled properly
debugged issue with fortran backend where file names that contained a directory path were not handled properly for module imports
debugged issue with fortran backend where adjustments of the default auto meta parameters were not applied correctly
the state variable indices and parameter names returned as the fourth and third return values of CircuitTemplate.get_run_func, respectively, now use the frontend variable names instead of the backend variable names
implemented a method CircuitIR.get_frontend_varname that returns the frontend variable name given a backend variable name
changed the theme of the readthedocs documentation website
added documentation for all supported backend functions
added documentation for dependencies and requirements
added documentation for YAML template structure to the documentation website
added documentation for mathematical syntax
added the changelog to the documentation website
__getitem__methods on all frontend template classes that allow for a less convoluted examination of the major properties of the template classes
added pytests that test these new features
users can now quickly access each node on
CircuitTemplate, each operator on
EdgeTemplate, and each variable on
added class for interactive grid search results visualization to utility
changed organization of the pandas DataFrames that
grid-searchreturns: Each different parameterization of the model appears only once in the
resultsDataFrame uses a full hierarchical column organization.
The pandas DataFrame returned by
CircuitTemplate.runuses a fully hierarchical column organization now: Every node hierarchy level is a separate level in the column index hierarchy.
minor docstring improvements
fixed bug in edge equation setup where a wrong index was provided to the target variable sometimes
fixed bug in variable updating that occurred for
numpy.ndarrayvariables where the
shapeattribute was an empty tuple
applied all changed to the gallery examples in the documentation
added generic method for state variable indexing to circuit.py that is used for all edge-related indexing operations now (replacing multiple, slightly different implementations at various places in circuit.py)
added an alternative compute graph class that can be used to generate function files that do not perform in-place manipulations of the vectorfield
dybut instead just create a new variable. This is relevant for gradient-based optimization.
improved the modularity of the
added a method
add_importto the backend that allows adding import statements to the top of a function file
added a backend function
concatenatethat can be used in equation strings now in order to combine vectorized variables
removed a bug where calling
clear_frontend_cachesdid not clear all IR caches properly
added support for models with vectorized state-variables
improved performance of edge operations
more detailed output about returned function arguments when calling
improved memory consumption during model initialization
complex-valued models use complex variable types for all variables and parameters now, to prevent type conversions
added a new method
CircuitTemplate.get_varthat allows users to access backend variables after calling
added automated reduction of vectorized constants, if all constants are identical
added possibility to pass iterables to
CircuitTemplate.update_var, thus allowing to update vectorized variables in one go
CircuitTemplate.add_edges_from_matrixsuch that only edges with non-zero weights are added to the
run-function generating method of ComputeGraph now returns the keys of the function arguments together with the arguments
implemented a method in CircuitTemplate that allows to get the indices of state variables within the system state vector
added different versions of the Izhikevich mean-field model (the dimensionless model, the biophysical model with distributed background currents, and the biophysical model with distributed spike thresholds)
improved documentation gallery examples (debugged equations, added images, added Izhikevich model references)
added Heun’s method as a new differential equation solver method
Heun’s method was integrated with all backends
a test was added that ensures correct functionality of Heun’s method
the usage of the method is demonstrated in the simulations gallery example
added hyperlinks to websites explaining the different numerical solvers in the gallery example
improved the backend implementation of choosing between different solvers (less code overlap between backends now)
added support for delayed differential equation (DDE) systems
past(y, tau)is now available for any backend that allows to evaluate a state variable
edges with discrete delays that are to be used in combination with an adaptive step-size solver are translated into
a gallery example was added that demonstrates how to interface the Python package
ddeintvia a DDE system generated by PyRates
the Julia backend received support for performing DDE simulations from within PyRates via its interface to
debugged latex equation error in Izhikevich model gallery example
bugfix in julia backend where a wrong file ending was provided
added new pytests for the izhikevich model, the python model definition interface and the CircuitIR translation
updated the readme
added a new QIF model template that includes conductance-based synapses
added gallery example for the izhikevich mean-field model
updated default parameterization of the izhikevich model
added a matlab backend (mainly for code generation, since simulations are very slow due to array conversion between numpy and matlab)
added a mean-field model of the Izhikevich neuron
small bug fixes
removed an issue of the fortran interface to Auto-07p that led to wrong function argument indices being generated
removed an issue with synaptic weights of -1 being converted to 1
removed a compatibility issue between old and new versions of the ‘to_yaml’ methods
added the natural logarithm ‘log’ as backend function
removed bug where vectorized circuits with multiple edges to the same target wre not resolved correctly
removed bug where creating deepcopies of a
CircuitTemplateraised an error for scalar-valued models
added a new gallery example demonstrating different ways of adding delays to models
added a new gallery example demonstrating the different options to optimize run times of numerical simulations
added support for complex-valued systems
added model templates for the kuramoto order parameter and the theta neuron model
added model templates for the van der pol oscillator and the stuart-landau oscillator
added support for Python 3.9
added new example galleries
extended pytest library
CircuitTemplate.to_yamlmethod that allows to save a given
CircuiTemplateinstance to a YAML definition file
CircuitTemplate.add_edges_from_matrixmethod that allows to connect nodes in a
CircuiTemplateinstance via connectivity matrices
deleted old, deprecated code fragments
removed the dependecy on pyparsing
updates to changelog and setup.py
new backends (torch, Julia)
sympy-based equation parsing
improved compute graph
improved generation of run functions from compute graphs
additional convenience functions for simulations
less steps from model definition to simulation
reduced syntax for model definitions
Removed utility package
utility packages for parameter optimization, signal analysis and visualization have been removed from the pyrates main package
most utility functionalities have been moved to separate repositories of the pyrates-neuroscience organization
less package requirements
new model templates
improved structure of the model templates
New model templates and documentation examples
new example galleries and jupyter notebooks with hands-on use examples
CircuitIR._add_edge_buffer()was re-worked, such that the algorithm that translates gamma-kernel convolutions for edges into ODE systems is more transparent and computationally less expensive
additionally improved the source code documentation of
removed unnecessary copying/indexing operations of original edge source variable
fixed a bug in
CircuitIR._add_edge_buffer()that caused a mix-up between edges when data was transferred from the originial output into the buffer variables.
zero-weight edges are now removed much earlier in the compilation process, thus reducing compilation time.
fixed a bug in
CircuitIR._add_edge_buffer()that caused a mix-up between edges when some outputs of a node had delays while others had not.
CircuitIR.vectorize_edges()in circuit.py such that zero-weight edges are removed during the vectorization, even if they have a delay defined on them (previously, defining a delay on a zero-weight edge kept that edge in the graph).
corrected mistake in the documentation of
pyrates.ir.circuit.CircuitIR.add_edge_buffer(), where arguments that refer to the source variable of an edge, where erroneously described as target variable information.
fixed bug in
pyrates.ir.circuit.CircuitIR.add_edge_buffer()where the conversion from discrete delays to gamma-kernel convolutions led to a mix-up between different edges in some special cases.
fixed bug in
pyrates.utility.pyauto.PyAuto._start_from_solution()where certain special solution branches from Auto-07p could not be properly handled
pyrates.utility.grid_search.adapt_circuit()such that node properties are always deep-copied before they are changed. This allows users to change the values of parameters on specific node operators, even though that exact same operator has been used to define multiple nodes in the network. Previously, changing the value of the parameter on one node led to changes on all other nodes as well.
improved stability and usability of
pyrates.utility.visualization.Interactive2DParamPlot. A title for the 2D plot can now be passed, a colorbar is added, and the location of the axis ticks of the 2D plot was improved
all Jansen-Rit model introductions where changed to track the excitatory and inhibitory post-synaptic potentials of the pyramidal cell population as output variables. Their difference provides the average membrane potential of the pyramidal cells.
Changed documentation jupyter notebooks etc. to account for Jansen-Rit model definition change (see below).
qif_fold.pyto delete all temporary files created by auto-07p
model templates updates:
added a 3 population model to the qif model templates in
added qif population template with mono-exponential synaptic depression to
added a new model template to
simple_montbrio.yamlwhich provides a QIF population with mono-exponential spike-frequency adaptation
added bi-exponential short-term adaptation descriptions to QIF models in
small change to the Jansen-Rit model definition: I removed the observer operator. To investigate the PC membrane potential, please record both PSP variables at the PC population and plot their sum. This has been changed accordingly in all corresponding examples.
PyAuto related updates:
pyrates.utility.pyauto.PyAuto.to_filemethod. Additional keyword arguments that are provided by the user are now stored in a dictionary under
additional_attributes. Loading a pyauto instance via
from_filewill thus create an attribute
additional_attributeson the instances, which will contain all the keyword arguments as a dictionary.
pyrates.utility.pyauto.get_from_solutionsmethod. Previously, providing more than one attribute key resulted in the method using an erroneous list comprehension style. This was fixed now. Providing multiple keys now results in the method returning a list of lists.
changed the way automatic re-runs of starting points computed by auto are detected by
fixed problem with extracting a solution from auto via the method
pyrates.utility.pyauto.PyAuto.get_solution(). Apparently, sometimes the function call
solution_branch(solution_key)does not work and throws an attribute error. I implemented a work around for this inconsistency in the Python interface for auto-07p.
pyrates.utility.pyauto.continue_period_doubling_bfto return a list that contains the names of all period doubling continuations performed with the pyauto instance that is returned as a second return value
now catching an error in the plotting-related method
pyrates.utility.pyauto.PyAuto._get_line_collection, if the
xargument is a vector of length 1
pyrates.utility.pyauto.PyAuto.get_point_idx(). Sometimes, when auto-07p failed to locate the new fixed point of a steady-state solution, it retries the previous step. PyAuto could not recognize the auto-07p diagnostic output for such cases. Now it can.
improved period doubling continuation in
pyrates.utility.pyauto.py. Only solution branches with new PD bifurcations are saved for plotting etc.
pyrates.utility.pyauto.PyAuto.plot_continuationmethod such that it can be used to plot continuations of the time parameter “PAR(14)”
pyrates.utility.pyauto.PyAuto.plot_trajectoryto be able to plot phase space trajectories of explicit time continuations (continuations in “PAR(14)”)
adjusted the return values of the
pyrates.utility.pyauto.fractal_dimensionmethod for its extreme cases. If the sum of the lyapunov spectrum is positive, return the number of lyapunov exponents. If the largest lyapunov exponent is smaller or equal to zero, use the normal formula.
cutoffargument to the
pyrates.utility.pyauto.PyAuto.plot_trajectorymethod that allows to cut off initial transients within the time window from
implemented speed-up of
pyrates.utility.pyauto.PyAuto.get_eigenvalues()method and fixed two bugs with the method that (1) led to an empty list being returned, and (2) caused the method to fail when applied to a steady-state solution
improved continuation of period doubling cascades via
pyrates.utility.pyauto.continue_period_doubling_bf(): It recognizes now which branches it had already switched to at period doubling bifurcations. Reduces the number of overall continuations
added the possibility to pass the installation directory of auto-07p to
pyrates.ir.circuit.CircuitIR.to_pyauto(). This makes it easier to install auto-07p, since the users do not have to manupilate system path variables themselfes anymore
debugged counting of already calculated parameter continuations in
pyrates.ir.circuit.CircuitIR.clear()method together with the
pyrates.backend.fortran_backend.FortranBackend.clear()method to remove all temporary files created by us or auto-07p during the model compilation and execution.
added a warning to the
pyrates.utility.grid_search.grid_search()function if a certain parameter is not found in the model
improved interface between
added a keyword argument
grid_searchthat prevents removal of temporary files if set to
improved the interactive 2D plot in
pyrates.utility.visualization.Interactive2DParamPlot: retrieving the column index of each column name now handles multi-column Dataframes correctly.
replaced “is” comparisons with “==” comparisons where appropriate
evolutionary optimization updates:
changed the way model ids are sampled in
pyrates.utility.genetic_algorithm.DifferentialEvolutionAlgorithm. With the old method, multiple workers sometimes generated models with equal IDs, leading to errors.
added an argument to
pyrates.utility.genetic_algorithm.DifferentialEvolutionAlgorithm.run()that allows to suppress runtime warnings.
intermediate representation updates:
fixed a bug in
pyrates.ir.circuit.CircuitIR._add_edge_buffer()that led to a wrong association between node indices and node variables in cases where multiple delayed edges with different delay profiles had to be handled. This mostly affected grid-searches over delay distribution parameters.
pyrates.ir.circuit.CircuitIR.run()to the backend run function. Now all printed output of PyRates can be muted.
Removed conversion function register, because the functions were not used and made the code unnecessarily complicated
might be replaced by a graph-based conversion path-finder in the future, if necessary
Extended support for loading circuits from and saving to files
supported classes: templates
Removed all imports in
pyrates.utility.__init__.pyfor increased stability. Previously, importing something from
pyrates.utility, would have required a user to install optional packages that might not have been needed. Now all utility functions need to be imported from sub-files in the
pyrates.utilitymodule instead of directly from the module.
Added optional install collection
teststhat includes all packages necessary to run the tests. Also restricted the travis CI build to use only the tests installation instead of the full installation.
Added feature to pass a dictionary to
CircuitTemplate.apply()in order to adapt values of variables on the fly. This behaviour was already supported by all other parts of the hierarchy, only circuits missed out until now.
Added experimental support for multiple source variables per edge
edges can either have multiple input variable from the same input node, or
they can have additional (“modulating”) input from any node in the network
Added experimental support for Fortran code creation backend
Edge delays can now be transformed into delay distributions via convoluted Gamma-Kernels based on differential equation using a mean and spread parameter for the delay
various performance improvements
0.8.2 Included bug fixes from jajcayn:¶
Allow to initialise CircuitTemplate with instances of
EdgeTemplateinstead of a template path, previous behaviour is unaffected.
Fix writing graph to the file by passing
_formatalong until the end
0.8.1 Improved cluster distribution and bug fixes¶
updated tensorflow dependency to >=2.0, fixes some dependency problems
Improved cluster distribution system, available under
New feature: model optimization with genetic algorithms, available under
Miscellaneous bug fixes
removed version ID numbers of operator/node instances in the intermediate representation. I.e. a node label
mynodewas previously renamed to
mynode.0and will now keep it’s original label.
moved all functionality of ComputeGraph into CircuitIR, which is now the main interface for the backend.
CircuitIRnow has a
.compilemethod that performs all vectorization and transformation into the computable backend form.
vectorization will transform all nodes into instances of
VectorizedNodeIRthat have labels like
vector_nodeXwith X being a integer index. The map between old nodes and vectorized nodes with respective index is saved in the
label_mapdictionary attribute of the
When adding input or sampling output of a network with multiple stacked levels of circuits, you can now use
allto get all nodes within that particular level. For example
mysubcircuit1/all/mynodewill get all nodes with label
mynodethat are in one level of sub-circuits below
Tensorflow support now relies on the current 2.0 release candidate
Added optional install requirements via