|
Results:
-
Extension of the Modeling Field
Theory neural network for the classification of objects (as
seen in ICANN 06)
the system is able to classify
multi-feature objects from complex stimulus set
Description:
-
Test 1: We have tested the
scaling-up of the multi-dimensional MFT algorithm with a complex
categorization data set. The training environment is composed of
1000 objects belonging to the following 10 2-feature object
prototypes: [0.1, 0.8], [0.2, 1.0], [0.3, 0.1], [0.4, 0.5],
[0.5, 0.2], [0.6, 0.3], [0.7, 0.4], [0.8, 0.9], [0.9, 0.6] and
[1.0, 0.7]. For each prototype, we generated 100 objects using a
Gaussian distribution with standard deviation of 0.05. During
training, we used 10 initial random fields.
Click for full size image
Figure 1: time evolution of
the 10 concept-models fields. The analysis of results also shows
the successful identification of the 10 prototype models and the
matching between the 100 stimuli generated by each object and
the final values of the fields.
Up -
Back
|
