This paper highlights the importance of modeling the geometry of linear features in order to generalize them. It proposes a constructive approach by segmenting lines and qualifying sections detected according to different criteria at several levels ("hierarchical" process). The aim of such an approach is clearly to guide the choices of adequate sequences of generalization operations and algorit...

We propose a simple, yet effective approach for spatiotemporal feature learning using deep 3-dimensional convolutional networks (3D ConvNets) trained on a large scale supervised video dataset. Our findings are three-fold: 1) 3D ConvNets are more suitable for spatiotemporal feature learning compared to 2D ConvNets; 2) A homogeneous architecture with small 3 × 3 × 3 convolution kernels in all lay...

Bi-linear feature learning models, like the gated autoencoder, were proposed as a way to model relationships between frames in a video. By minimizing reconstruction error of one frame, given the previous frame, these models learn “mapping units” that encode the transformations inherent in a sequence, and thereby learn to encode motion. In this work we extend bi-linear models by introducing “hig...

Recent research on growth empirics has been focused on resolving model and variable uncertainty. The conventional approach has been to assume a linear growth process and then to proceed with investigating the relevant variables that determine cross-country growth. This paper questions the linearity assumption underlying the vast majority of such research and uses recently-developed nonparametri...

we present a steganalysis scheme for lsb matching steganography based on feature vectors extracted from integer wavelet transform (iwt). in integer wavelet decomposition of an image, the coefficients will be integer, so we can calculate co-occurrence matrix of them without rounding the coefficients. before calculation of co-occurrence matrices, we clip some of the most significant bitplanes of ...

the problem of automatic signature recognition has received little attention incomparison with the problem of signature verification, despite its potentialapplications for many business processes and can be used effectively in paperlessoffice projects. this paper presents model-based off-line signature recognition withrotation invariant features. non-linear rotation of signature patterns is one...

Nonparametric methods for the estimation of the link function in generalized linear models are able to avoid bias in the regression parameters. But for the estimation of the link typically the full model, which includes all predictors, has been used. When the number of predictors is large these methods fail since the full model can not be estimated. In the present article a boosting type method...

We present a method for learning higher-order polynomial functions from examples using linear regression and feature construction. Regression is used on a set of training instances to produce a weight vector for a linear function over the feature set. If this hypothesis is imperfect, a new feature is constructed by forming the product of the two features that most eeectively predict the squared...

We have sets X and Y: we will assume that Y is a finite set. Our goal is to build a model that estimates the conditional probability p(y|x) of a label y ∈ Y given an input x ∈ X . For example, x might be a word, and y might be a candidate partof-speech (noun, verb, preposition etc.) for that word. We have a feature-vector definition φ : X × Y → Rd. We also assume a parameter vector w ∈ Rd. Give...