Package topo :: Package transferfn

Package transferfn

A family of function objects for transforming a matrix generated from some other function.

Output functions are useful for neuron activation functions, normalization of matrices, etc. They come in two varieties: OutputFunction, and CFPOutputFunction. An OutputFunction (e.g. PiecewiseLinear) applies to one matrix of any type, such as an activity matrix or a set of weights. To apply an OutputFunction to all of the weight matrices in an entire CFProjection, an OutputFunction can be plugged in to CFPOF_Plugin. CFPOF_Plugin is one example of a CFPOutputFunction, which is a function that works with the entire Projection at once.

Any new output functions added to this directory will automatically become available for any model.

$Id: __init__.py 9831 2009-01-25 16:34:35Z ceball $

Version: $Revision: 9831 $

Submodules

[hide private]

topo.transferfn.basic: Simple functions operating on a matrix, potentially modifying it.
topo.transferfn.misc: Transfer functions with more complex dependencies.
topo.transferfn.optimized: Output functions (see basic.py) and projection-level output functions (see projfn.py) written in C to optimize performance.
topo.transferfn.projfn: Output functions (see basic.py) that apply to a whole Projection.

Classes

[hide private]

ActivityAveragingTF
Calculates the average of the input activity.

BinaryThreshold
Forces all values below a threshold to zero, and above it to 1.0.

DivisiveNormalizeL1
TransferFn that divides an array by its L1 norm.

DivisiveNormalizeL2
TransferFn to divide an array by its Euclidean length (aka its L2 norm).

DivisiveNormalizeLinf
TransferFn to divide an array by its L-infinity norm (i.e.

DivisiveNormalizeLp
TransferFn to divide an array by its Lp-Norm, where p is specified.

ExpLinear
Transfer function that is exponential until t from which point it is linear.

GeneralizedLogistic
The generalized logistic curve (Richards' curve): y = l + (u /(1 + b * exp(-r*(x-2*m))^(1/b))).

HalfRectifyAndPower
Transfer function that applies a half-wave rectification (i.e., clips at zero), and then raises the result to the e-th power (where the exponent e can be selected arbitrarily).

HalfRectifyAndSquare
Transfer function that applies a half-wave rectification (clips at zero) and then squares the values.

HomeostaticMaxEnt
Implementation of homeostatic intrinsic plasticity from Jochen Triesch, ICANN 2005, LNCS 3696 pp.65-70.

Hysteresis
Smoothly interpolates a matrix between simulation time steps, with exponential falloff.

IdentityTF
Identity function, returning its argument as-is.

NakaRushton
Naka-Rushton curve.

PiecewiseLinear
Piecewise-linear TransferFn with lower and upper thresholds.

PoissonSample
Simulate Poisson-distributed activity with specified mean values.

ScalingTF
Scales input activity based on the current average activity (x_avg).

Sigmoid
Sigmoidal (logistic) transfer function: 1/(1+exp-(r*x+k)).

Square
Transfer function that applies a squaring nonlinearity.

Threshold
Forces all values below a threshold to zero, and leaves others unchanged.

TransferFn
Function object to modify a matrix in place, e.g.

TransferFnWithRandomState
Abstract base class for TransferFns that use a random number generator.

TransferFnWithState
Abstract base class for TransferFns that need to maintain a self.plastic parameter.

Variables

[hide private]

__package__ = 'topo.transferfn'