Skip to main content
. Author manuscript; available in PMC: 2012 May 1.
Published in final edited form as: Cogn Sci. 2011 May;35(4):638–681. doi: 10.1111/j.1551-6709.2011.01174.x

Figure 1.

Figure 1

Architecture of the Serbian morphology model. Ellipses represent groups of units organized into layers. Arrows indicate directional weighted connections used to pass information among layers. The input layer is connected to the phonological layer via two hidden layers consisting of 250 and 100 units, respectively. The second hidden layer, which directly feeds into the output layer, is connected in both directions to a ‘recurrent’ layer consisting of 100 units. The inclusion of this layer permitted the network to produce sequential outputs when presented with a static input.