Our main contribution in this paper is a new reduction from explicit two-source extractors for polynomially-small entropy rate and negligible error to explicit t-non-malleable extractors with seed-length that has a good dependence on t. Our reduction is based on the Chattopadhyay and Zuckerman framework [CZ16], and surprisingly we dispense with the use of resilient functions which appeared to b...

We introduce the notion of a super-strong extractor. Given two independent weak random sources X,Y , such extractor EXT(·, ·) has the property that EXT(X,Y ) is statistically random even if one is given Y . Namely, 〈Y, EXT(X,Y )〉 ≈ 〈Y,R〉. Super-strong extractors generalize the notion of strong extractors [16], which assume that Y is truly random, and extractors from two weak random sources [26,...

Discovering knowledge from textual sources and subsequently expanding the coverage of knowledge bases like DBPedia or Freebase currently requires either extensive manual work or carefully designed information extractors. Information extractors capture triples from textual sentences. Each triple consists of a subject, a predicate/property, and an object. Triples can be mediated via verbs, nouns,...

This article describes a method of classifying webpages using features extracted from the design. Design features are language independent and correspond well to a user’s first impression of a page. A number of new feature extractors were developed, capable of extracting human-readable features. These features are based on guidelines for use of color and composition in web design. The compositi...

With the increasing number of personal albums and photos in them, users have more and more problems with their organization [4]. This is due to lack of descriptive data, however, their amount only depends on the abilities and will of the image owners. Tools for metadata creation are available – the main problem is with the user motivation: because tagging and annotating of photos is usually a b...

Music tag words that describe music audio by text have different levels of abstraction. Taking this issue into account, we propose a music classification approach that aggregates multilevel and multi-scale features using pre-trained feature extractors. In particular, the feature extractors are trained in sample-level deep convolutional neural networks using raw waveforms. We show that this appr...

I will survey some of the recent exciting progress on explicit constructions of randomness extractors for independent sources. Many of the new constructions rely on explicit constructions of newly introduced pseudorandom primitives, and there remains scope of finding better explicit constructions of these primitives. I will also discuss some possible approaches for constructing optimal Ramsey g...

This paper presents and evaluates two linear feature extractors based on mutual information. These feature extractors consider general dependencies between features and class labels, as opposed to statistical techniques such as PCA which does not consider class labels and LDA, which uses only simple rst order dependencies. As evidenced by several simulations on high dimensional data sets, the p...