nn.modules.RateJax

class nn.modules.RateJax(*args, **kwargs)[source]

Bases: rockpool.nn.modules.jax.jax_module.JaxModule

Encapsulates a population of rate neurons, supporting feed-forward and recurrent modules, with a Jax backend

Examples

Instantiate a feed-forward module with 8 neurons:

>>> mod = RateJax(8,)
RateEulerJax 'None' with shape (8,)

Instantiate a recurrent module with 12 neurons:

>>> mod_rec = RateJax(12, has_rec = True)
RateEulerJax 'None' with shape (12,)

Instantiate a feed-forward module with defined time constants:

>>> mod = RateJax(7, tau = np.arange(7,) * 10e-3)
RateEulerJax 'None' with shape (7,)

This module implements the update equations:

\[ \begin{align}\begin{aligned}\dot{X} = -X + i(t) + W_{rec} H(X) + bias + \sigma \zeta_t X = X + \dot{x} * dt / au\\H(x, t) = relu(x, t) = (x - t) * ((x - t) > 0)\end{aligned}\end{align} \]

Attributes overview

`class_name`	Class name of `self`
`full_name`	The full name of this module (class plus module name)
`name`	The name of this module, or an empty string if `None`
`shape`	The shape of this module
`size`	(DEPRECATED) The output size of this module
`size_in`	The input size of this module
`size_out`	The output size of this module
`spiking_input`	If `True`, this module receives spiking input.
`spiking_output`	If `True`, this module sends spiking output.
`rng_key`	The Jax PRNG key for this module
`x`	A vector `(N,)` of the internal state of each unit
`w_rec`	The recurrent weight matrix `(N, N)` for this module
`tau`	The vector `(N,)` of time constants \(\tau\) for each unit
`bias`	The vector `(N,)` of bias currents for each unit
`threshold`	(Tensor) Unit thresholds `(Nout,)` or `()`
`dt`	The Euler solver time step for this module
`noise_std`	The std.
`act_fn`	(Callable) Activation function

Methods overview

`__init__`(shape[, tau, bias, threshold, ...])	Instantiate a non-spiking rate module, either feed-forward or recurrent.
`as_graph`()	Convert this module to a computational graph
`attributes_named`(name)	Search for attributes of this or submodules by time
`evolve`(input_data[, record])	Evolve the state of this module over input data
`modules`()	Return a dictionary of all sub-modules of this module
`parameters`([family])	Return a nested dictionary of module and submodule Parameters
`reset_parameters`()	Reset all parameters in this module
`reset_state`()	Reset the state of this module
`set_attributes`(new_attributes)	Assign new attributes to this module and submodules
`simulation_parameters`([family])	Return a nested dictionary of module and submodule SimulationParameters
`state`([family])	Return a nested dictionary of module and submodule States
`timed`([output_num, dt, add_events])	Convert this module to a `TimedModule`
`tree_flatten`()	Flatten this module tree for Jax
`tree_unflatten`(aux_data, children)	Unflatten a tree of modules from Jax to Rockpool

__init__(shape: typing.Union[int, typing.Tuple[jax._src.numpy.lax_numpy.ndarray]], tau: typing.Optional[typing.Union[float, numpy.ndarray, torch.Tensor]] = None, bias: typing.Optional[typing.Union[float, numpy.ndarray, torch.Tensor]] = None, threshold: typing.Optional[typing.Union[float, numpy.ndarray, torch.Tensor]] = None, w_rec: typing.Optional[jax._src.numpy.lax_numpy.ndarray] = None, weight_init_func: typing.Callable = <function unit_eigs>, has_rec: bool = False, activation_func: typing.Union[str, typing.Callable] = <function H_ReLU>, noise_std: float = 0.0, dt: float = 0.001, rng_key: typing.Optional[int] = None, *args: list, **kwargs: dict)[source]

Instantiate a non-spiking rate module, either feed-forward or recurrent.

Parameters

shape (Tuple[np.ndarray]) – A tuple containing the shape of this module. If one dimension is provided (N,), it will define the number of neurons in a feed-forward layer. If two dimensions are provided, a recurrent layer will be defined. In that case the two dimensions must be identical (N, N).
tau (float) – A scalar or vector defining the initialisation time constants for the module. If a vector is provided, it must match the output size of the module. Default: 20ms
bias (float) – A scalar or vector defining the initialisation bias values for the module. If a vector is provided, it must match the output size of the module. Default: 0.
w_rec (np.ndarray) – An optional matrix defining the initialisation recurrent weights for the module. Default: Normal / sqrt(N)
has_rec (bool) – Iff True, the module operates in recurrent mode. Default: False, operate in feed-forward mode.
weight_init_func (Callable) – A function used to initialise the recurrent weights, if used. Default: unit_eigs(); initialise such that recurrent feedback has eigenvalues distributed within the unit circle.
activation_func (Callable) – The activation function of the neurons. This can be provided as a string ['ReLU', 'sigmoid', 'tanh'], or as a function that accepts a vector of neural states and returns the vector of output activations. This function must use jax.numpy math functions, and not numpy math functions. Default: 'ReLU'.
dt (float) – The Euler solver time-step. Default: 1e-3
noise_std (float) – The std. dev. of normally-distributed noise added to the neural state at each time step. Default: 0.
rng_key (Any) – A Jax PRNG key to initialise the module with. Default: not provided, the module PRNG will be initialised with a random number.
*args – Additional positional arguments
**kwargs – Additional keyword arguments

_abc_impl = <_abc_data object>

_auto_batch(data: jax._src.numpy.lax_numpy.ndarray, states: Tuple = (), target_shapes: Optional[Tuple] = None) → Tuple[jax._src.numpy.lax_numpy.ndarray, Tuple[jax._src.numpy.lax_numpy.ndarray]]

Automatically replicate states over batches and verify input dimensions

Usage:

>>> data, (state0, state1, state2) = self._auto_batch(data, (self.state0, self.state1, self.state2))

This will verify that data has the correct final dimension (i.e. self.size_in). If data has only two dimensions (T, Nin), then it will be augmented to (1, T, Nin). The individual states will be replicated out from shape (a, b, c, ...) to (n_batches, a, b, c, ...) and returned.

Parameters

data (np.ndarray) – Input data tensor. Either (batches, T, Nin) or (T, Nin)
states (Tuple) – Tuple of state variables. Each will be replicated out over batches by prepending a batch dimension

Returns

(np.ndarray, Tuple[np.ndarray]) data, states

_force_set_attributes: (bool) If True, do not sanity-check attributes when setting.

_get_attribute_family(type_name: str, family: Optional[Union[Tuple, List, str]] = None) → dict

Search for attributes of this module and submodules that match a given family

This method can be used to conveniently get all weights for a network; or all time constants; or any other family of parameters. Parameter families are defined simply by a string: "weights" for weights; "taus" for time constants, etc. These strings are arbitrary, but if you follow the conventions then future developers will thank you (that includes you in six month’s time).

Parameters

type_name (str) – The class of parameters to search for. Must be one of ["Parameter", "SimulationParameter", "State"] or another future subclass of ParameterBase
family (Union[str, Tuple[str]]) – A string or list or tuple of strings, that define one or more attribute families to search for

Returns

A nested dictionary of attributes that match the provided type_name and family

Return type

dict

_get_attribute_registry() → Tuple[Dict, Dict]

Return or initialise the attribute registry for this module

Returns: registered_attributes, registered_modules
Return type: (tuple)

_has_registered_attribute(name: str) → bool

Check if the module has a registered attribute

Parameters: name (str) – The name of the attribute to check
Returns: True if the attribute name is in the attribute registry, False otherwise.
Return type: bool

_in_Module_init: (bool) If exists and True, indicates that the module is in the __init__ chain.

_name: Optional[str]: Name of this module, if assigned

_register_attribute(name: str, val: rockpool.parameters.ParameterBase)

Record an attribute in the attribute registry

Parameters

name (str) – The name of the attribute to register
val (ParameterBase) – The ParameterBase subclass object to register. e.g. Parameter, SimulationParameter or State.

_register_module(name: str, mod)

Add a submodule to the module registry

Parameters

name (str) – The name of the submodule, extracted from the assigned attribute name
mod (JaxModule) – The submodule to register

_reset_attribute(name: str) → rockpool.nn.modules.module.ModuleBase

Reset an attribute to its initialisation value

Parameters: name (str) – The name of the attribute to reset
Returns: For compatibility with the functional API
Return type: self (Module)

_rockpool_pytree_registry = []: The internal registry of registered JaxModule s

_shape: The shape of this module

_spiking_input: bool: Whether this module receives spiking input

_spiking_output: bool: Whether this module produces spiking output

_submodulenames: List[str]: Registry of sub-module names

_wrap_recorded_state(recorded_dict: dict, t_start: float) → Dict[str, rockpool.timeseries.TimeSeries]

Convert a recorded dictionary to a TimeSeries representation

This method is optional, and is provided to make the timed() conversion to a TimedModule work better. You should override this method in your custom Module, to wrap each element of your recorded state dictionary as a TimeSeries

Parameters

state_dict (dict) – A recorded state dictionary as returned by evolve()
t_start (float) – The initial time of the recorded state, to use as the starting point of the time series

Returns

The mapped recorded state dictionary, wrapped as TimeSeries objects

Return type

Dict[str, TimeSeries]

act_fn: P_Callable: (Callable) Activation function

as_graph() → rockpool.graph.graph_base.GraphModuleBase[source]

Convert this module to a computational graph

Returns: The computational graph corresponding to this module
Return type: GraphModuleBase
Raises: NotImplementedError – If as_graph() is not implemented for this subclass

attributes_named(name: Union[Tuple[str], List[str], str]) → dict

Search for attributes of this or submodules by time

Parameters: name (Union[str, Tuple[str]) – The name of the attribute to search for
Returns: A nested dictionary of attributes that match name
Return type: dict

bias: P_ndarray: The vector (N,) of bias currents for each unit

property class_name: str

Class name of self

Type: str

dt: P_float: The Euler solver time step for this module

evolve(input_data: jax._src.numpy.lax_numpy.ndarray, record: bool = False)[source]

Evolve the state of this module over input data

NOTE: THIS MODULE CLASS DOES NOT PROVIDE DOCUMENTATION FOR ITS EVOLVE METHOD. PLEASE UPDATE THE DOCUMENTATION FOR THIS MODULE.

Parameters

input_data – The input data with shape (T, size_in) to evolve with
record (bool) – If True, the module should record internal state during evolution and return the record. If False, no recording is required. Default: False.

Returns

(output, new_state, record): output (np.ndarray): The output response of this module with shape (T, size_out) new_state (dict): A dictionary containing the updated state of this and all submodules after evolution record (dict): A dictionary containing recorded state of this and all submodules, if requested using the record argument

Return type

tuple

property full_name: str

The full name of this module (class plus module name)

Type: str

modules() → Dict

Return a dictionary of all sub-modules of this module

Returns: A dictionary containing all sub-modules. Each item will be named with the sub-module name.
Return type: dict

property name: str

The name of this module, or an empty string if None

Type: str

noise_std: P_float: The std. dev. \(\sigma\) of noise added to internal neuron states at each time step

parameters(family: Optional[Union[Tuple, List, str]] = None) → Dict

Return a nested dictionary of module and submodule Parameters

Use this method to inspect the Parameters from this and all submodules. The optional argument family allows you to search for Parameters in a particular family — for example "weights" for all weights of this module and nested submodules.

Although the family argument is an arbitrary string, reasonable choises are "weights", "taus" for time constants, "biases" for biases…

Examples

Obtain a dictionary of all Parameters for this module (including submodules):

>>> mod.parameters()
dict{ ... }

Obtain a dictionary of Parameters from a particular family:

>>> mod.parameters("weights")
dict{ ... }

Parameters: family (str) – The family of Parameters to search for. Default: None; return all parameters.
Returns: A nested dictionary of Parameters of this module and all submodules
Return type: dict

reset_parameters()

Reset all parameters in this module

Returns: The updated module is returned for compatibility with the functional API
Return type: Module

reset_state() → rockpool.nn.modules.jax.jax_module.JaxModule

Reset the state of this module

Returns: The updated module is returned for compatibility with the functional API
Return type: Module

rng_key: Union[np.ndarray, State]: The Jax PRNG key for this module

set_attributes(new_attributes: Union[collections.abc.Iterable, collections.abc.MutableMapping, collections.abc.Mapping]) → rockpool.nn.modules.jax.jax_module.JaxModule

Assign new attributes to this module and submodules

Parameters: new_attributes (Tree) – The tree of new attributes to assign to this module tree
Return type: JaxModule

property shape: tuple

The shape of this module

Type: tuple

simulation_parameters(family: Optional[Union[Tuple, List, str]] = None) → Dict

Return a nested dictionary of module and submodule SimulationParameters

Use this method to inspect the SimulationParameters from this and all submodules. The optional argument family allows you to search for SimulationParameters in a particular family.

Examples

Obtain a dictionary of all SimulationParameters for this module (including submodules):

>>> mod.simulation_parameters()
dict{ ... }

Parameters: family (str) – The family of SimulationParameters to search for. Default: None; return all SimulationParameter attributes.
Returns: A nested dictionary of SimulationParameters of this module and all submodules
Return type: dict

property size: int

(DEPRECATED) The output size of this module

Type: int

property size_in: int

The input size of this module

Type: int

property size_out: int

The output size of this module

Type: int

property spiking_input: bool

If True, this module receives spiking input. If False, this module expects continuous input.

Type: bool

property spiking_output

If True, this module sends spiking output. If False, this module sends continuous output.

Type: bool

state(family: Optional[Union[Tuple, List, str]] = None) → Dict

Return a nested dictionary of module and submodule States

Use this method to inspect the States from this and all submodules. The optional argument family allows you to search for States in a particular family.

Examples

Obtain a dictionary of all States for this module (including submodules):

>>> mod.state()
dict{ ... }

Parameters: family (str) – The family of States to search for. Default: None; return all State attributes.
Returns: A nested dictionary of States of this module and all submodules
Return type: dict

tau: P_ndarray: The vector (N,) of time constants \(\tau\) for each unit

threshold: P_ndarray: (Tensor) Unit thresholds (Nout,) or ()

timed(output_num: int = 0, dt: Optional[float] = None, add_events: bool = False)

Convert this module to a TimedModule

Parameters

output_num (int) – Specify which output of the module to take, if the module returns multiple output series. Default: 0, take the first (or only) output.
dt (float) – Used to provide a time-step for this module, if the module does not already have one. If self already defines a time-step, then self.dt will be used. Default: None
add_events (bool) – Iff True, the TimedModule will add events occurring on a single timestep on input and output. Default: False, don’t add time steps.

Returns: TimedModule: A timed module that wraps this module

tree_flatten() → Tuple[tuple, tuple]: Flatten this module tree for Jax

classmethod tree_unflatten(aux_data, children): Unflatten a tree of modules from Jax to Rockpool

w_rec: P_ndarray: The recurrent weight matrix (N, N) for this module

x: P_ndarray: A vector (N,) of the internal state of each unit