Computational Intelligence and Neuroscience / 2009 / Article / Fig 5

Research Article

Unsupervised Learning of Overlapping Image Components Using Divisive Input Modulation

Figure 5

Node activations generated in response to input images containing overlapping πš— πš– 𝚏 𝚜 𝚎 𝚚 pixel square components. (a) Test images. Numbers indicate which components are present in each test image. (b)–(f) Images reconstructed from the response of each network. Numbers indicate which components are represented by strongly active nodes in each network (nodes that have an output activation greater than the mean of all node activations). Each network was given predefined weights so that distinct nodes represented 𝚍 πš’ πš– pixel squares at all possible locations.
381457.fig.005a
(a) Test images
381457.fig.005b
(b) 𝑠 = 2
381457.fig.005c
(c) 𝑠 = 3
381457.fig.005d
(d) 𝑠 = 4
381457.fig.005e
(e) 𝐩 = [ 0 . 1 , 0 . 1 ]
381457.fig.005f
(f) 𝐜 = [ 1 , 1 ]

Article of the Year Award: Outstanding research contributions of 2020, as selected by our Chief Editors. Read the winning articles.