Name of the inventor: Tony Jebara
Vectorized Input Data Encoding, Transmitting, and Decoding: Data representation, invariance learning and modeling of the represented data are described.
The system and method for compressing data represented by vectors comprises a set of sample data. Each vector is associated with a variable permutation operator that permutes the ordering of the pixels in the vector. The data in set is associated with a convex cost function which estimates the cost of permuting. The range of allowed permutations is linearly constrained. The cost function is statistically defined as a determinant of a covariance of the data. A processor minimizes the convex cost function over the constrained range of allowed permutations to identify a linear subspace of the bag of pixels. A principal components analyzer identifies a set of eigenvectors that span the linear subspace. An encoder compresses data by using the set of eigenvectors as a basis set to encode the data.
Data Compression Using Vectors
The system and method provide techniques to describe the raw data with a reduced number of variables, and thereby allowing volumes of raw data to be advantageously.