|
Kategorie: Diplomové, bakalářské práce |
Tento dokument chci!
Práce popisuje základy principu funkčnosti neuronů a vytvoření umělých neuronových sítí. Je zde důkladně popsána struktura a funkce neuronů a ukázán nejpoužívanější algoritmus pro učení neuronů. Základy fuzzy logiky, včetně jejich výhod a nevýhod, jsou rovněž prezentovány. Detailněji je popsán algoritmus zpětného šíření chyb a adaptivní neuro-fuzzy inferenční systém. Tyto techniky poskytují efektivní způsoby učení neuronových sítí.
Thanks its capability and because can perform the same function, almost
any neural network can replaced ANFIS.
(1997) class adaptive networks that are functionally equivalent fuzzy
inference systems (FIS), where the parameters fuzzy inference systems are
updated neural networks from set training data. ADAPTIVE NEURO FUZZY INFERENCE SYSTEM
Adaptive Neuro Fuzzy Inference System (ANFIS) developed Jang al.1 Learning algorithm ANFIS
The standard ANFIS uses the Sugeno-type fuzzy model generate fuzzy
rules from given input-output data set. adaptive network, its
name implies, network structure consisting nodes and directional links through
which the nodes are connected. For easy understanding, let's take simple
version fuzzy inference system with two inputs and one output rule set for
. The hybrid learning
algorithm includes two stages, which are:
forward pass identifies the consequent parameters with the help FIS
learning mechanism and least-squares estimator (LSE)
backward pass propagates backward the error rates (error backpropagation)
and updates the premise parameters the gradient descent method
In ANFIS, the membership functions (gaussian functions) are expected map
all inputs changing their parameters.
ANFIS consists self-tuning Sugeno-type inference system and calculates
its outputs weighted linear combination the consequents.
The fuzzy membership function the basic block fuzzy logic systems and
has many possible interpretations [10].22
5. Moreover, part all the nodes are adaptive,
which means their outputs depend the parameters pertaining these nodes, and
the learning rule specifies how these parameters should changed minimize
a prescribed error measure. desired that all inputs can mapped to
produce the desired outputs. Unfortunately, the case that there occur variations in
the inputs, the desired outputs will poorly approximated the actual outputs
because limitations finding the parameters the fixed finite number fuzzy
membership functions [10]. can define the richness the extracted
information from the given data case highly nonlinear systems and the form of
the membership functions can extended cover this richness. ANFIS enjoys many the advantages claimed neural
networks (NNs) and the linguistic interpretability fuzzy inference systems, wherein
both NNs and FIS play active roles effort reach specific goals [10], [11].
5. Its primary advantages are non-
linearity and structural knowledge representation
