Information Technology Reference
In-Depth Information
Fig. 4. Training data. The mean values of the two weight conditions (light and heavy, top) and
the four visual conditions (matching symbols, bottom) are shown. These mean time series are
used as prototypes for training the RNNPB. Vertical gray shaded areas represents the up and
down movement, whereas back and forth movements are unshaded. The area surrounding the
signals delineates two standard deviations from the mean.
and used to compute the visual prototype for circular-shaped objects. To find the pro-
prioceptive prototype for e.g. all heavy objects, all individual measurements with this
property ( n =40 ) are aggregated and used to calculate the mean value at each time
step. The subclass prototypes are then combined to form a 2-D multi-modal time series
that serves as an input for the recurrent neural network during training.
Search WWH ::




Custom Search