Learning Features and Predictive Transformation Encoding Based on a Horizontal Product Model

The visual system processes the features and movement of an object in separate pathways, called the ventral and dorsal streams. To integrate this principle in a functional model, a recurrent predictive network with a horizontal product is introduced. Learned in an unsupervised manner, two sets of hidden units represent cells in the ventral and dorsal pathways, respectively. Experiments show that the activity in the ventral-like units persists, given that the same feature appears in the receptive field, whilst the activity in the dorsal-like units shows a fluctuating pattern with different directions of object movements. Moreover, we show that the position information predicts the input’s future position taking into account its moving direction due to the direction-selective responses of the dorsal-like units.

@InProceedings{ZWW12a, 
 	 author =  {Zhong, Junpei and Weber, Volker and Wermter, Stefan},  
 	 title = {Learning Features and Predictive Transformation Encoding Based on a Horizontal Product Model}, 
 	 booktitle = {Proceedings of the 22nd International Conference on Artificial Neural Networks (ICANN 2012)},
 	 editors = {},
 	 number = {},
 	 volume = {},
 	 pages = {539--546},
 	 year = {2012},
 	 month = {Sep},
 	 publisher = {},
 	 doi = {10.1007/978-3-642-33269-2_68}, 
 }