+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ Abstract base classes for defining SOMs.
+ + + +AbstractSom
+
+
+
+ Bases: Module
Abstract base class for SOM models.
+ +src/somap/core.py50 + 51 + 52 + 53 + 54 + 55 + 56 + 57 + 58 + 59 + 60 + 61 + 62 + 63 + 64 + 65 + 66 + 67 + 68 + 69 + 70 + 71 + 72 + 73 + 74 + 75 + 76 + 77 + 78 + 79 + 80 + 81 + 82 + 83 + 84 + 85 + 86 + 87 + 88 + 89 + 90 + 91 + 92 + 93 + 94 + 95 + 96 + 97 + 98 + 99 +100 +101 +102 +103 +104 +105 +106 +107 +108 +109 +110 +111 +112 +113 +114 +115 +116 +117 +118 +119 +120 +121 +122 +123 +124 +125 +126 +127 +128 +129 +130 +131 +132 +133 +134 +135 +136 +137 +138 +139 +140 +141 +142 +143 +144 +145 +146 +147 +148 +149 +150 +151 +152 +153 +154 +155 +156 +157 +158 +159 +160 +161 +162 +163 +164 +165 +166 +167 +168 +169 +170 +171 +172 +173 +174 +175 +176 +177 +178 +179 +180 +181 +182 +183 +184 +185 +186 +187 +188 +189 +190 +191 +192 +193 +194 +195 +196 +197 +198 +199 +200 +201 +202 +203 +204 | |
__call__(input)
+
+Makes a single iteration.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
input |
+
+ InputData
+ |
+
+
+
+ Data array for the given SOM models. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
| + | +
+
+
+ A tuple with the new SOM model and the auxilary data. + |
+
src/somap/core.py115 +116 +117 +118 +119 +120 +121 +122 +123 +124 +125 +126 +127 +128 +129 +130 +131 +132 +133 +134 +135 +136 +137 +138 +139 +140 +141 +142 +143 +144 +145 +146 +147 +148 +149 +150 +151 +152 +153 +154 +155 +156 +157 +158 +159 +160 +161 +162 +163 +164 +165 +166 +167 +168 +169 +170 +171 +172 +173 +174 +175 +176 | |
__init__(shape, topography, borderless, input_shape, params, metrics=True, debug=False, key=jax.random.PRNGKey(0))
+
+Creates a SOM models.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
shape |
+ + | +
+
+
+ Shape of the 2D map. + |
+ + required + | +
topography |
+ + | +
+
+
+ Topography of the 2D map. Either 'square' for a square grid +or 'hex' for hexagonal grid. + |
+ + required + | +
borderless |
+ + | +
+
+
+ Toroidal topography if True, meaning that the top (resp. left) +border meets the bottom (resp. right) border. + |
+ + required + | +
input_shape |
+ + | +
+
+
+ Shape of the input data. + |
+ + required + | +
params |
+ + | +
+
+
+ Parameters of the SOM (depends on the SOM flavor). + |
+ + required + | +
metrics |
+ + | +
+
+
+ If True, returns quantization and topographic errors as auxilary +data. + |
+
+ True
+ |
+
debug |
+ + | +
+
+
+ If True, returns debug data as auxilary data. + |
+
+ False
+ |
+
key |
+
+ PRNGKeyArray
+ |
+
+
+
+ JAX random key used during map initialization. + |
+
+ PRNGKey(0)
+ |
+
src/somap/core.pybulk_set(attr_dict)
+
+Sets multiples attributes at once.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
attr_dict |
+ + | +
+
+
+ dictionary where keys are attribute names and values are +attributes values to be set. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
| + | +
+
+
+ A new instance of the updated object. + |
+
src/somap/core.pygenerate_algo(params)
+
+Converts specific SOM parameters into generic SOM functions.
+ + +set(attribute, value)
+
+Sets an attribute to a specific value.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
attribute |
+
+ str
+ |
+
+
+
+ name of the attribute. + |
+ + required + | +
value |
+ + | +
+
+
+ new value of the attribute. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
| + | +
+
+
+ A new instance of the updated object. + |
+
src/somap/core.pyAbstractSomParams
+
+
+InputData
+
+
+
+ Bases: TypedDict
Structure of the input data.
+ +Classical SOMs only have the 'bu_v' bottum-up input value. +Other inputs allow to create more complex SOMs receiving top-down and lateral +inputs with a mask.
+TypedDict instead of dataclass to facilitate future modifications.
+src/somap/core.pySomAlgo
+
+
+make_step(model, input)
+
+Makes a single iteration.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
model |
+
+ AbstractSom
+ |
+
+
+
+ SOM model. + |
+ + required + | +
input |
+
+ InputData
+ |
+
+
+
+ Data array for the given SOM models. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
| + | +
+
+
+ A tuple with the new SOM model and the auxilary data. + |
+
src/somap/core.pymake_steps(model, inputs)
+
+Makes multiple iterations at once.
+Uses the jax.lax.scan() function to optimize computations.
Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
model |
+
+ AbstractSom
+ |
+
+
+
+ SOM model. + |
+ + required + | +
inputs |
+ + | +
+
+
+ Batch data array for the given SOM models. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
| + | +
+
+
+ A tuple with the new SOM model and the auxilary data. + |
+
src/somap/core.pyCatalog of distance functions.
+dist means distance
+wdist means weighted distance
All distance functions are defined for comparison between 2 1D vectors.
+ + + +AbstractDist
+
+
+
+ Bases: Module
Ensures that all distance functions have the same signatures.
+ +src/somap/distance.py__call__(a, b)
+
+
+ abstractmethod
+
+
+SOM distance function.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
a |
+
+ Float[Array, ' n']
+ |
+
+
+
+ 1D array. + |
+ + required + | +
b |
+
+ Float[Array, ' n']
+ |
+
+
+
+ 1D array. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
+ Float[Array, '']
+ |
+
+
+
+ The distance as a scalar. + |
+
src/somap/distance.pyCyclicEuclidianDist
+
+
+
+ Bases: AbstractDist
Cyclic euclidian distance function.
+ +src/somap/distance.py__call__(a, b)
+
+Returns a cyclic euclidian distance where 0 and 1 are equals.
+More optimized than the arccos(cos()) based cyclic distance.
+Takes advantage of data type's limits.
src/somap/distance.pyEuclidianDist
+
+
+
+ Bases: AbstractDist
Euclidian distance function.
+ +src/somap/distance.py__call__(a, b)
+
+dist_cim(v1, v2, sigma)
+
+The Correntropy Induced Metric (CIM) distance.
+It is based on correntropy which is a generalized correlation. +It computes the distance between two vectors without suffering from the curse of +dimensionality like the Euclidian distance thanks to a kernel function. +The chosen kernel is a Gaussian function (parameterised by 'sigma'). +For large sigma, the CIM distance tends to the L2 metric. +For very small sigma, the CIM distance tends to the L0 norm.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
v1 |
+ + | +
+
+
+ first vector to be compared + |
+ + required + | +
v2 |
+ + | +
+
+
+ second vector to be compared + |
+ + required + | +
sigma |
+ + | +
+
+
+ kernel bandwidth + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
| + | +
+
+
+ A scalar value between 0 and 1 corresponding to the similarity between v1 and v2 + |
+
Raises:
+| Type | +Description | +
|---|---|
+ ValueError
+ |
+
+
+
+ incompatible shapes of v1 and v2 + |
+
src/somap/distance.pywdist_l2(a, b, w)
+
+Computes a weighted version of the euclidian norm.
+ +src/somap/distance.pyCatalog of learning rate functions.
+ + + +AbstractLr
+
+
+
+ Bases: Module
Ensures that all learning rate functions have the same signatures.
+ +src/somap/learning_rate.py__call__(t, distances)
+
+
+ abstractmethod
+
+
+SOM learning rate function.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
t |
+
+ Integer[Array, '']
+ |
+
+
+
+ Current iteration. + |
+ + required + | +
distances |
+
+ Float[Array, 'x y']
+ |
+
+
+
+ Distances between the prototype weights and the input data. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
+ Float[Array, ''] | Float[Array, 'x y']
+ |
+
+
+
+ A scalar ar 2D array containing the learning rate. + |
+
src/somap/learning_rate.pyConstantLr
+
+
+
+ Bases: AbstractLr
Basic SOM learning rate function.
+ +src/somap/learning_rate.pyDsomLr
+
+
+
+ Bases: AbstractLr
DSOM learning rate function.
+ +src/somap/learning_rate.py__call__(_, distances)
+
+KsomLr
+
+
+
+ Bases: AbstractLr
Kohonen SOM learning rate function.
+ +src/somap/learning_rate.py__call__(t, _)
+
+Catalog of neighborhood functions.
+Neighborhood functions are defined as equinox.Module parametrized functions
AbstractNbh
+
+
+
+ Bases: Module
Ensures that all neighborhood functions have the same signatures.
+ +src/somap/neighborhood.py__call__(distance_map, t, quantization_error)
+
+
+ abstractmethod
+
+
+SOM Neighborhood function.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
distance_map |
+
+ Float[Array, 'x y']
+ |
+
+
+
+ Distance of each grid elements from the winning element. + |
+ + required + | +
t |
+
+ Integer[Array, '']
+ |
+
+
+
+ Current iteration. + |
+ + required + | +
quantization_error |
+
+ Float[Array, '']
+ |
+
+
+
+ The computed difference between the winner prototype +and the input. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
+ Float[Array, 'x y']
+ |
+
+
+
+ The neighborhood distance. + |
+
src/somap/neighborhood.pyDsomNbh
+
+
+
+ Bases: AbstractNbh
Dynamic Kohonen neighborhood function.
+ +src/somap/neighborhood.py__call__(distance_map, _, quantization_error)
+
+Computes the Dynamic SOM neighboring value of each grid element.
+ +Nicolas P. Rougier, Yann Boniface. Dynamic Self-Organising Map. +Neurocomputing, Elsevier, 2011, 74 (11), pp.1840-1847. +ff10.1016/j.neucom.2010.06.034ff. ffinria-00495827
+Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
self |
+ + | +
+
+
+ self.plasticity: Dynamic value to compute the neighbourhood distance. + |
+ + required + | +
distance_map |
+
+ Float[Array, 'x y']
+ |
+
+
+
+ Distance of each element from the winner element. + |
+ + required + | +
_ |
+ + | +
+
+
+ Not used + |
+ + required + | +
quantization_error |
+
+ Float[Array, '']
+ |
+
+
+
+ The computed difference between the winner prototype +and the input. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
+ Float[Array, 'x y']
+ |
+
+
+
+ The neighborhood distance, as calculated in the article. + |
+
src/somap/neighborhood.pyGaussianNbh
+
+
+
+ Bases: AbstractNbh
Exponentially decreasing neighborhood function.
+ +src/somap/neighborhood.py__call__(distance_map, t, __)
+
+Return the Kohonen time-independent neighboring value of each element.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
self |
+ + | +
+
+
+ Module's parameters +self.sigma: Neighbourhood distance. + |
+ + required + | +
distance_map |
+
+ Float[Array, 'x y']
+ |
+
+
+
+ Distance of each element from the winner element. + |
+ + required + | +
t |
+ + | +
+
+
+ Not used + |
+ + required + | +
__ |
+ + | +
+
+
+ Not used + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
+ Float[Array, 'x y']
+ |
+
+
+
+ The kohonen neighborhood distance. + |
+
src/somap/neighborhood.pyKsomNbh
+
+
+
+ Bases: AbstractNbh
Kohonen neighborhood function.
+ +src/somap/neighborhood.py__call__(distance_map, t, _)
+
+Returns the Kohonen neighboring value of each element.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
self |
+ + | +
+
+
+ Module's parameters +self.t_f: Aimed iteration. +self.sigma_i: Current neighborhood distance. +self.sigma_f: Aimed neighborhood distance. + |
+ + required + | +
distance_map |
+
+ Float[Array, 'x y']
+ |
+
+
+
+ Distance of each grid elements from the winning element. + |
+ + required + | +
t |
+
+ Integer[Array, '']
+ |
+
+
+
+ Current iteration. + |
+ + required + | +
_ |
+ + | +
+
+
+ Not used + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
+ Float[Array, 'x y']
+ |
+
+
+
+ The kohonen neighborhood distance. + |
+
src/somap/neighborhood.pyCatalog of different flavors of SOMs.
+To add a new SOMs, define the 2 following classes: +
+ + + +Dsom
+
+
+
+ Bases: AbstractSom
Dynamic SOM.
+ +src/somap/som.pygenerate_algo(p)
+
+
+ staticmethod
+
+
+Converts Dynamic SOM parameters into generic SOM functions.
+ +src/somap/som.pyDsomParams
+
+
+
+ Bases: AbstractSomParams
Dynamic SOM parameters.
+ +src/somap/som.pyKsom
+
+
+
+ Bases: AbstractSom
Kohonen SOM.
+ +src/somap/som.pygenerate_algo(p)
+
+
+ staticmethod
+
+
+Converts Kohonen SOM parameters into generic SOM functions.
+ +src/somap/som.pyKsomParams
+
+
+
+ Bases: AbstractSomParams
Kohonen SOM parameters.
+ +src/somap/som.pySom
+
+
+
+ Bases: AbstractSom
Generic SOM.
+Distance, neighborhood, learning rate and update functions are directly specified. +This is the more flexible way of defining a SOM.
+ +src/somap/som.pygenerate_algo(p)
+
+
+ staticmethod
+
+
+Identity function returning generic SOM functions.
+ + +SomParams
+
+
+
+ Bases: AbstractSomParams, SomAlgo
Generic SOM parameters.
+Same as the SomAlgo class but with the AbstractSomParams parent.
StaticKsom
+
+
+
+ Bases: AbstractSom
Time-independant Kohonen SOM.
+ +src/somap/som.pygenerate_algo(p)
+
+
+ staticmethod
+
+
+Converts Static Kohonen SOM parameters into generic SOM functions.
+ +src/somap/som.pyStaticKsomParams
+
+
+
+ Bases: AbstractSomParams
Time-independant Kohonen SOM parameters.
+ +src/somap/som.pyCatalog of update functions.
+ + + +AbstractUpdate
+
+
+
+ Bases: Module
Ensures that all update functions have the same signatures.
+ +src/somap/update.py__call__(lr, nbh, input_bu, w_bu)
+
+
+ abstractmethod
+
+
+SOM Update function.
+ + + +Parameters:
+| Name | +Type | +Description | +Default | +
|---|---|---|---|
lr |
+ + | +
+
+
+ Learning rate. + |
+ + required + | +
nbh |
+ + | +
+
+
+ Neighborhood. + |
+ + required + | +
input_bu |
+ + | +
+
+
+ Data input. + |
+ + required + | +
w_bu |
+ + | +
+
+
+ Prototype weights. + |
+ + required + | +
Returns:
+| Type | +Description | +
|---|---|
| + | +
+
+
+ The updated prototype weights. + |
+
src/somap/update.pyCyclicSomUpdate
+
+
+
+ Bases: AbstractUpdate
Cyclic update functions.
+ +src/somap/update.py__call__(lr, nbh, input_bu, w_bu)
+
+Updates the prototype weights where 0 is the same as 1.
+ +src/somap/update.pySomUpdate
+
+
+
+ Bases: AbstractUpdate
Generic update function.
+ +src/somap/update.py__call__(lr, nbh, input_bu, w_bu)
+
+Updates the prototype weights.
+ +src/somap/update.py