Wyniki wyszukiwania - Biblioteka Nauki

Ten serwis zostanie wyłączony 2025-02-11.

Nowa wersja platformy, zawierająca wyłącznie zasoby pełnotekstowe, jest już dostępna.
Przejdź na https://bibliotekanauki.pl

Ograniczanie wyników

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: Jaccard similarity

Sortuj według:

Ogranicz wyniki do:

Analysis of Compounds Activity Concept Learned by SVM Using Robust Jaccard Based Low-dimensional Embedding

100%

Jastrzębski S. , Czarnecki W. M.

Schedae Informaticae

2015

tom Vol. 24

9--19

Support Vector Machines (SVM) with RBF kernel is one of the most successful models in machine learning based compounds biological activity prediction. Unfortunately, existing datasets are highly skewed and hard to analyze. During our research we try to answer the question how deep is activity concept modeled by SVM. We perform analysis using a model which embeds compounds’ representations in a low-dimensional real space using near neighbour search with Jaccard similarity. As a result we show that concepts learned by SVM is not much more complex than slightly richer nearest neighbours search. As an additional result, we propose a classification technique, based on Locally Sensitive ashing approximating the Jaccard similarity through minhashing technique, which performs well on 80 tested datasets (consisting of 10 proteins with 8 different representations) while in the same time allows fast classification and efficient online training.

Extensions for Apple-Google exposure notification mechanism

44%

Bobowski A. , Cichoń J. , Kutyłowski M.

tom Vol. 69, nr 4

art. no. e137126

We analyze the Google-Apple exposure notification mechanism designed by the Apple-Google consortium and deployed on a large number of Corona-warn apps. At the time of designing it, the most important issue was time-to-market and strict compliance with the privacy protection rules of GDPR. This resulted in a plain but elegant scheme with a high level of privacy protection. In this paper we go into details and propose some extensions of the original design addressing practical issues. Firstly, we point to the danger of a malicious cryptographic random number generator (CRNG) and resulting possibility of unrestricted user tracing. We propose an update that enables verification of unlinkability of pseudonymous identifiers directly by the user. Secondly, we show how to solve the problem of verifying the “same household” situation justifying exempts from distancing rules. We present a solution with MIN-sketches based on rolling proximity identifiers from the Apple-Google scheme. Thirdly, we examine the strategies for revealing temporary exposure keys. We have detected some unexpected phenomena regarding the number of keys for unbalanced binary trees of a small size. These observations may be used in case that the size of the lists of diagnosis keys has to be optimized.