|
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.
However,
fuzzy system applications are limited the fields where expert knowledge is
available and the number input variables small.
The main reason behind the creation intelligent hybrid systems have been
these limitations.18
behavior based system input-output data. The
. short comparison between the
operation fuzzy systems and neural networks presented the following table:
Skills Fuzzy Systems Neural Nets
Knowledge
acquisition
Inputs Human experts Sample sets
Tools Interaction Algorithms
Uncertainty
Information Quantitive and Qualitive Quantitive
Cognition Decision making Perception
Reasoning
Mechanism Heuristic search Parallel computat.
To overcome the problem knowledge acquisition, neural networks are
extended automatically extract fuzzy rules from numerical data [6]. Fuzzy logic systems are good in
decision explanations but the rules they use make those decisions they cannot
acquire automatically.
Speed Low High
Adaptation
Fault-tolerance Low Very high
Learning Induction Adjusting weights
Natural
language
Implementation Explicit Implicit
Flexibility High Low
Table 4.) making them suited for individual problems.
Every intelligent technique has some computational qualities (explanation of
decisions, learning ability, etc. This learned knowledge can used to
generate fuzzy logic rules and membership functions, significantly reducing the
development time. This provides more cost effective solution fuzzy
implementation typically less expensive alternative than neural nets for
embedded control applications. there complex application
with two different sub-problems, then neural network and expert system can be
used separately for solving these individual tasks. Expressing the weights the neural net using fuzzy
rules helps provide greater insights into the neural nets, thus leading design of
better neural nets [5].1 Properties fuzzy systems and neural networks
(based [6])
Neural network learning techniques can automate the process design and
tune the membership functions and reduce the development time and cost in
a large measure. For
example, while neural networks are good recognizing patterns, they are not good
at explaining how they reach their decisions [6]. The behavior fuzzy systems can explained with the help of
fuzzy rules and their performance can adjusted tuning the rules. With the combination two more techniques, able to
overcome the limitations individual techniques
