|
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.
9 The results the NNV1 network tested with words Speaker1
.40
8.2 Running the simulation
After the program was made and its adequate functionality was tested, the
next step experimentation with and fine tuning the simulation parameters for
optimal results.
8.1.1 Network parameters
For the training and the testing process the following parameters were set within the
NNV network:
Number layers: 3
Output function the neuron: sigmoid function
Training function: error backpropagation
Number epochs: 2000
Threshold: the biggest output the three networks indicates
the recognized language
Number outputs: 3
The training was done the function netlearn while the testing done by
neteval.1.
8.1 The 'NNV' network
This section was created using the mentioned algorithms Chapter 3.
a) The network tested with words Speaker1
Actual language Czech English Hungarian ∑
All words 153
Precisely identified words 100
Efficiency 56,86% 78,43% 60,78% 65,36%
Imprecisely identified words
(instead the actual language)
as
Hungarian
as
English
as
Hungarian
as
Czech
as
English
as
Czech
∑
15 53
Ratio 29,41% 13,73% 15,69% 5,88% 19,61% 19,61% 34,64%
Table 8. total number trains and tests were run with two speakers
(Speaker1 and Speaker2) allocate the average error rate while the analysis
parameters were set params.m follows:
framestep=20; %ms
framelen=25; %ms
melfilerbankcount=10;
With this setting, one simulation took approximately 700 seconds
