Jacobs, R. A., and Kosslyn, S. M., 1994, Encoding shape and spatial relations: The role of receptive field size in coordinating complementary representations.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....the retina and 12 for the category names (Fig. 2) The same type and number of units are used in the output layer. The 49 (7x7) retina input units consists of gaussian receptive field units. These process the 50x50 pixels of the original image, each using a square receptive field of 11x11 pixel [1,4]. Retina units are divided into two groups, the periphery and the center. The 6 hidden units are also divided into two groups of 3 units each, one specialized for shapes and one for textures. The periphery input units send connections only to the 3 shape hidden units. The retina central units send ....

Jacobs RA, Kosslyn SM (1994). Encoding shape and spatial relations: The role of receptive field size in coordinating complementary representations. Cognitive Science, 18: 361-386.

....halves is the clearest fact about its anatomy, but the processing implications of this division are far from clear. Several researchers have developed models that reflect this division. Cohen, Romero, Farah and Servan Schreiber [2] have modelled opponent attentional processors. Jacobs and Kosslyn [3] have explored categorical and coordinate processing of spatial input. Reggia, Goodall and Shkuro [4] have studied lateralisation in reading and letter recognition tasks. Elsewhere we have presented a variety of split models of neglect and of visual lexical processing [5,6,7,8] all of which ....

Jacobs, R.A. & Kosslyn. S.M. (1994). Encoding shape and spatial relations: the role of receptive field size in coordinating complementary representations. Cognitive Science, 18, 361--386.

....learning task this should not be confused with techniques for unsupervised competitive learning in neural networks. Dailey Cottrell 6 Figure 3: Example face, book, cup, and can images Jacobs (1997) argues, they have also been very useful tools for exploring hypotheses about brain function. Jacobs and Kosslyn (1994), for instance, showed that if one expert in a two expert network was endowed with large receptive fields and the other was given smaller receptive fields, one expert specialized for a what task whereas the other specialized for a where task. As another example, Erickson and Kruschke (1998) ....

Jacobs, R. A. and Kosslyn, S. M. (1994). Encoding shape and spatial relations --- The role of receptive field size in coordinating complementary representations. Cognitive Science, 18(3):361--386.

....model. Other research that has been conducted in this area includes: the application of the mixtureof experts model to control strategies (Jacobs Jordan, 1993) the use of an ExpectationMaximisation algorithm for adjusting the parameters of a mixture of experts model (Jordan Jacobs, 1994); investigation of models comprising expert networks which vary in topology (Jacobs Kosslyn, 1994; Jacobs, 1997b) and, the examination of the bias and variance of these models (Jacobs, 1997a) 1.1 Mixture of experts model Typically the architecture of a mixture of experts model comprises n ....

.... application of the mixtureof experts model to control strategies (Jacobs Jordan, 1993) the use of an ExpectationMaximisation algorithm for adjusting the parameters of a mixture of experts model (Jordan Jacobs, 1994) investigation of models comprising expert networks which vary in topology (Jacobs Kosslyn, 1994; Jacobs, 1997b) and, the examination of the bias and variance of these models (Jacobs, 1997a) 1.1 Mixture of experts model Typically the architecture of a mixture of experts model comprises n expert networks and a single gating network (see Figure 1) The expert networks are expected to ....

Jacobs, R. A. & Kosslyn, S. M. (1994) Encoding shape and spatial relations: The role of receptive field size in coordinating complementary representations. Cognitive Science, 18, 361-386.

....face and a high pass filtered image of another individual s face, subjects consistently use the low frequency component of the image for the task. This work indicates that low spatial frequency information may be more important for face identification than high spatial frequency information. Jacobs and Kosslyn (1994) showed how differential availability of large and small receptive field sizes in a mixture of experts network (Jacobs, Jordan, Nowlan, and Hinton, 1991) can lead to experts that specialize for what and where tasks. In previous work, we proposed that a neural mechanism allocating resources ....

Jacobs, R. and Kosslyn, S. (1994). Encoding shape and spatial relations --- The role of receptive field size in coordinating complementary representations. Cognitive Science, 18(3):361--386.

....model. Other research that has been conducted in this area includes; the application of the mixtureof experts model to control strategies (Jacobs Jordan, 1993) the use of an ExpectationMaximisation algorithm for adjusting the parameters of a mixture of experts model (Jordan Jacobs, 1994); investigation of models comprising expert networks which vary in topology (Jacobs Kosslyn, 1994; Jacobs, 1997b) and, the examination of the bias and variance of these models (Jacobs, 1997a) Architecture The architecture of a mixture of experts model comprises n expert networks and a single ....

.... application of the mixtureof experts model to control strategies (Jacobs Jordan, 1993) the use of an ExpectationMaximisation algorithm for adjusting the parameters of a mixture of experts model (Jordan Jacobs, 1994) investigation of models comprising expert networks which vary in topology (Jacobs Kosslyn, 1994; Jacobs, 1997b) and, the examination of the bias and variance of these models (Jacobs, 1997a) Architecture The architecture of a mixture of experts model comprises n expert networks and a single gating network. An example of such a model, with two expert networks, is shown in Figure 2.7. The ....

Jacobs, R. A. & Kosslyn, S. M. (1994) Encoding shape and spatial relations: The role of receptive field size in coordinating complementary representations. Cognitive Science, 18, 361-386.

....low resolution pathway involving units with large receptive fields will be better able to accomplish the discrimination required in the face recognition task, whereas a highresolution pathway involving units with small receptive fields will be better able to accomplish the object recognition task. Jacobs and Kosslyn (1994) have successfully applied this approach within the mixture of experts paradigm. The hypothesis that face processing primarily depends on holistic or configural information, whereas processing other object types depends more on analyzing an object s subparts, is also testable in our model. We plan ....

Jacobs, R.A., & Kosslyn, S.M. (1994). Encoding shape and spatial relations: The role of receptive field size in coordinating complementary representations. Cognitive Science, 18(3), 361--386.

No context found.

Jacobs, R. A., and Kosslyn, S. M., 1994, Encoding shape and spatial relations: The role of receptive field size in coordinating complementary representations.

No context found.

Jacobs R A, and Kosslyn SM, 1994, "Encoding shape and spatial relations: The role of receptive field size in coordinating complemetary relations", Cog Science, 18, 361-386.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC