Limits...
Self-Organization of Spatio-Temporal Hierarchy via Learning of Dynamic Visual Image Patterns on Action Sequences.

Jung M, Hwang J, Tani J - PLoS ONE (2015)

Bottom Line: Up to now, however, most biological and computational modeling studies have mainly focused on the spatial domain and do not discuss temporal domain processing of the visual cortex.This model is characterized by the application of both spatial and temporal constraints on local neural activities, resulting in the self-organization of a spatio-temporal hierarchy necessary for the recognition of complex dynamic visual image patterns.Furthermore, an evaluation test for the recognition of concatenated sequences of those prototypical movement patterns indicated that the model is endowed with a remarkable capability for the contextual recognition of long-range dynamic visual image patterns.

View Article: PubMed Central - PubMed

Affiliation: Department of Electrical Engineering, KAIST, Daejeon, Republic of Korea.

ABSTRACT
It is well known that the visual cortex efficiently processes high-dimensional spatial information by using a hierarchical structure. Recently, computational models that were inspired by the spatial hierarchy of the visual cortex have shown remarkable performance in image recognition. Up to now, however, most biological and computational modeling studies have mainly focused on the spatial domain and do not discuss temporal domain processing of the visual cortex. Several studies on the visual cortex and other brain areas associated with motor control support that the brain also uses its hierarchical structure as a processing mechanism for temporal information. Based on the success of previous computational models using spatial hierarchy and temporal hierarchy observed in the brain, the current report introduces a novel neural network model for the recognition of dynamic visual image patterns based solely on the learning of exemplars. This model is characterized by the application of both spatial and temporal constraints on local neural activities, resulting in the self-organization of a spatio-temporal hierarchy necessary for the recognition of complex dynamic visual image patterns. The evaluation with the Weizmann dataset in recognition of a set of prototypical human movement patterns showed that the proposed model is significantly robust in recognizing dynamically occluded visual patterns compared to other baseline models. Furthermore, an evaluation test for the recognition of concatenated sequences of those prototypical movement patterns indicated that the model is endowed with a remarkable capability for the contextual recognition of long-range dynamic visual image patterns.

No MeSH data available.


Related in: MedlinePlus

Snapshot of the internal dynamics of (A) layer 5 and (B) layer 3 plotted on the two dimensional PCA space.(A) PCA results on the internal dynamics at the end of all the concatenated actions (time step = 99) in layer 5. Nine groups corresponding to the nine different concatenated action categories appeared in the plot. Those nine groups presented themselves as three large clusters arranged according to the second action categories (blue dotted triangle), and each large cluster was also sub-divided into three isomorphic sub-clusters according to the first action categories (red dotted triangle). (B) PCA results on the internal dynamics at the end of the second action of all the concatenated actions (time step = 84) in layer 3. Three clusters were organized based on the current action category.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4492609&req=5

pone.0131214.g006: Snapshot of the internal dynamics of (A) layer 5 and (B) layer 3 plotted on the two dimensional PCA space.(A) PCA results on the internal dynamics at the end of all the concatenated actions (time step = 99) in layer 5. Nine groups corresponding to the nine different concatenated action categories appeared in the plot. Those nine groups presented themselves as three large clusters arranged according to the second action categories (blue dotted triangle), and each large cluster was also sub-divided into three isomorphic sub-clusters according to the first action categories (red dotted triangle). (B) PCA results on the internal dynamics at the end of the second action of all the concatenated actions (time step = 84) in layer 3. Three clusters were organized based on the current action category.

Mentions: The functional hierarchy developed in the network was examined with principal component analysis (PCA) results to measure the activation states in the slow layer at the end of all the concatenated actions. When plotted, nine groups corresponding to the nine concatenated action categories emerged, with the groups arranged according to the first and last actions. Three large clusters (dependent on the second action) and three sub-clusters (dependent on the first action) are represented by blue and red triangles, respectively, in Fig 6A. Regardless of the scale, the bottom left, top center, and bottom right vertices of each triangle correspond with JP, OH, and TH, respectively. Interestingly, with the same PCA for the activation states in the middle layer at the end of the second action of all the concatenated actions, this type of self-similar structure across different scales was not observed in Fig 6B. Only three clusters appeared in the plot grouped merely on the current action category, regardless of the previous action.


Self-Organization of Spatio-Temporal Hierarchy via Learning of Dynamic Visual Image Patterns on Action Sequences.

Jung M, Hwang J, Tani J - PLoS ONE (2015)

Snapshot of the internal dynamics of (A) layer 5 and (B) layer 3 plotted on the two dimensional PCA space.(A) PCA results on the internal dynamics at the end of all the concatenated actions (time step = 99) in layer 5. Nine groups corresponding to the nine different concatenated action categories appeared in the plot. Those nine groups presented themselves as three large clusters arranged according to the second action categories (blue dotted triangle), and each large cluster was also sub-divided into three isomorphic sub-clusters according to the first action categories (red dotted triangle). (B) PCA results on the internal dynamics at the end of the second action of all the concatenated actions (time step = 84) in layer 3. Three clusters were organized based on the current action category.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4492609&req=5

pone.0131214.g006: Snapshot of the internal dynamics of (A) layer 5 and (B) layer 3 plotted on the two dimensional PCA space.(A) PCA results on the internal dynamics at the end of all the concatenated actions (time step = 99) in layer 5. Nine groups corresponding to the nine different concatenated action categories appeared in the plot. Those nine groups presented themselves as three large clusters arranged according to the second action categories (blue dotted triangle), and each large cluster was also sub-divided into three isomorphic sub-clusters according to the first action categories (red dotted triangle). (B) PCA results on the internal dynamics at the end of the second action of all the concatenated actions (time step = 84) in layer 3. Three clusters were organized based on the current action category.
Mentions: The functional hierarchy developed in the network was examined with principal component analysis (PCA) results to measure the activation states in the slow layer at the end of all the concatenated actions. When plotted, nine groups corresponding to the nine concatenated action categories emerged, with the groups arranged according to the first and last actions. Three large clusters (dependent on the second action) and three sub-clusters (dependent on the first action) are represented by blue and red triangles, respectively, in Fig 6A. Regardless of the scale, the bottom left, top center, and bottom right vertices of each triangle correspond with JP, OH, and TH, respectively. Interestingly, with the same PCA for the activation states in the middle layer at the end of the second action of all the concatenated actions, this type of self-similar structure across different scales was not observed in Fig 6B. Only three clusters appeared in the plot grouped merely on the current action category, regardless of the previous action.

Bottom Line: Up to now, however, most biological and computational modeling studies have mainly focused on the spatial domain and do not discuss temporal domain processing of the visual cortex.This model is characterized by the application of both spatial and temporal constraints on local neural activities, resulting in the self-organization of a spatio-temporal hierarchy necessary for the recognition of complex dynamic visual image patterns.Furthermore, an evaluation test for the recognition of concatenated sequences of those prototypical movement patterns indicated that the model is endowed with a remarkable capability for the contextual recognition of long-range dynamic visual image patterns.

View Article: PubMed Central - PubMed

Affiliation: Department of Electrical Engineering, KAIST, Daejeon, Republic of Korea.

ABSTRACT
It is well known that the visual cortex efficiently processes high-dimensional spatial information by using a hierarchical structure. Recently, computational models that were inspired by the spatial hierarchy of the visual cortex have shown remarkable performance in image recognition. Up to now, however, most biological and computational modeling studies have mainly focused on the spatial domain and do not discuss temporal domain processing of the visual cortex. Several studies on the visual cortex and other brain areas associated with motor control support that the brain also uses its hierarchical structure as a processing mechanism for temporal information. Based on the success of previous computational models using spatial hierarchy and temporal hierarchy observed in the brain, the current report introduces a novel neural network model for the recognition of dynamic visual image patterns based solely on the learning of exemplars. This model is characterized by the application of both spatial and temporal constraints on local neural activities, resulting in the self-organization of a spatio-temporal hierarchy necessary for the recognition of complex dynamic visual image patterns. The evaluation with the Weizmann dataset in recognition of a set of prototypical human movement patterns showed that the proposed model is significantly robust in recognizing dynamically occluded visual patterns compared to other baseline models. Furthermore, an evaluation test for the recognition of concatenated sequences of those prototypical movement patterns indicated that the model is endowed with a remarkable capability for the contextual recognition of long-range dynamic visual image patterns.

No MeSH data available.


Related in: MedlinePlus