arXiv subjects

Ofer Shapira

Publications and source records attributed to Ofer Shapira.

5 recordsLinked to original sources

arXiv2022

Improving Precancerous Case Characterization via Transformer-based Ensemble Learning

The application of natural language processing (NLP) to cancer pathology reports has been focused on detecting cancer cases, largely ignoring precancerous cases. Improving the characterization of precancerous adenomas assists in developing diagnostic tests for early cancer detection and prevention, especially for colorectal cancer (CRC). Here we developed transformer-based deep neural network NLP models to perform the CRC phenotyping, with the goal of extracting precancerous lesion attributes and distinguishing cancer and precancerous cases. We achieved 0.914 macro-F1 scores for classifying patients into negative, non-advanced adenoma, advanced adenoma and CRC. We further improved the performance to 0.923 using an ensemble of classifiers for cancer status classification and lesion size named entity recognition (NER). Our results demonstrated the potential of using NLP to leverage real-world health record data to facilitate the development of diagnostic tests for early cancer prevention.

cs.LG↗

arXiv2015

Polar Codes with Mixed-Kernels

A generalization of the polar coding scheme called mixed-kernels is introduced. This generalization exploits several homogeneous kernels over alphabets of different sizes. An asymptotic analysis of the proposed scheme shows that its polarization properties are strongly related to the ones of the constituent kernels. Simulation of finite length instances of the scheme indicate their advantages both in error correction performance and complexity compared to the known polar coding structures.

cs.IT↗

arXiv2015

Binary Polarization Kernels from Code Decompositions

In this paper, code decompositions (a.k.a. code nestings) are used to design binary polarization kernels. The proposed kernels are in general non-linear. They provide a better polarization exponent than the previously known kernels of the same dimensions. In particular, non-linear kernels of dimensions 14, 15, and 16 are constructed and are shown to have optimal asymptotic error-correction performance. The optimality is proved by showing that the exponents of these kernels achieve a new upper bound that is developed in this paper.

cs.IT↗

arXiv2011

Binary Polar Code Kernels from Code Decompositions

Code decompositions (a.k.a code nestings) are used to design good binary polar code kernels. The proposed kernels are in general non-linear and show a better rate of polarization under successive cancelation decoding, than the ones suggested by Korada et al., for the same kernel dimensions. In particular, kernels of sizes 14, 15 and 16 are constructed and shown to provide polarization rates better than any binary kernel of such sizes.

cs.IT↗

arXiv2005

Localization of light in a random grating array in a single mode fiber

We investigate light propagation in randomly spaced fiber gratings in a single mode fiber, and demonstrate the localization effect. Localization of light in random media resembles that of electrons in disordered solids, resulting from a subtle wave interference formation. We measured the light transmission after each additional grating fabrication and found an exponential decay that follows the localization theory. An important feature of the random array is its similarity to ordered gratings in the transmission and reflection behavior at the long array regime. Besides the basic interest in localization in one-dimensional systems, random grating arrays have potential applications, utilizing the possibility to fabricate long structures with strong and broadband reflections.

physics.optics↗