Search

The brain has a remarkable ability to recognize objects from noisy or corrupted sensory inputs. How this cognitive robustness is achieved computationally remains unknown. We present a coding paradigm, which encodes structural dependence among features of the input and transforms various forms of the same input into the same representation. The paradigm, through dimensionally expanded representation and sparsity constraint, allows redundant feature coding to enhance robustness and is efficient in representing objects. We demonstrate consistent representations of visual and olfactory objects under conditions of occlusion, high noise or with corrupted coding units. Robust face recognition is achievable without deep layers or large training sets. The paradigm produces both complex and simple receptive fields depending on learning experience, thereby offers a unifying framework of sensory processing. One line abstract We present a framework of efficient coding of objects as a combination of structurally dependent feature groups that is robust against noise and corruption.

Differences in expression level among different clades of VRs. (A) Expression of all V1r clades are represented in all strains. Clade J receptor genes are more highly expressed than receptors of other clades. (B) Expression of all V2r clades are represented in all strains. (PDF 558Â kb)

Sequence comparison of functionalized pseudogene. Alignment of Vmn1r-ps27 from SWR with Vmn1r-ps27 and Vmn1r42 from B6. (PDF 814Â kb)

Polymorphism and differential expression of FPR genes. A) Fpr-rs3. B) Fpr3. (PDF 2426Â kb)

Differentially expressed chemosensory genes other than V1r and V2r families. (PDF 138Â kb)

Strain-specific expression of vomeronasal receptors. Data is displayed as expressed (blue) or not expressed (white) to highlight the exclusive patterns of expression for some of the genes. (PDF 87Â kb)

High degree of polymorphism and differential expression of VR genes among strains. (A-F) Example track files illustrating the mapping of reads to individual VR genes. Each track is a superposition of four individual samples with SNPs highlighted as vertical lines with substitutions represented as follows: thymine as red, guanine as brown, cytosine as blue, and adenine as green. (PDF 4541Â kb)

Distribution of uniquely mapped, multi-mapped and unmapped reads among the samples presented as total reads (A) and percentage of reads (B). (PDF 464Â kb)