Unconventional computing devices operating on nonlinear chemical media offer an interesting alternative to standard, semiconductor-based computers. In this work we consider database classifiers formed of interacting droplets in which a photosensitive variant of Belousov-Zhabotinsky (BZ) reaction proceeds. We introduce an evolutionary algorithm that searches for optimal construction of a droplets-based classifier for a given problem. The algorithm is based on maximizing the mutual information between the database and the observed evolution of medium. As an example application of chemical database classifiers we apply the idea to the dataset of points belonging to a unit cube. The dataset contains two output classes: 1 for points belonging to a sphere with radius 0.5 located in the cube center, and 0 for points outside of the sphere. The reliability of optimized chemical classifiers of such database for different numbers of droplets involved in data processing is presented.
The present paper1 aims to propose a new type of information-theoretic method to maximize mutual information between inputs and outputs. The importance of mutual information in neural networks is well known, but the actual implementation of mutual information maximization has been quite difficult to undertake. In addition, mutual information has not extensively been used in neural networks, meaning that its applicability is very limited. To overcome the shortcoming of mutual information maximization, we present it here in a very simplified manner by supposing that mutual information is already maximized before learning, or at least at the beginning of learning. The method was applied to three data sets (crab data set, wholesale data set, and human resources data set) and examined in terms of generalization performance and connection weights. The results showed that by disentangling connection weights, maximizing mutual information made it possible to explicitly interpret the relations between inputs and outputs.
3
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Increasing application of non-invasive medical techniques (like stereotactic radiosurgery) generates a high demand for modern image processing algorithms. Image registration and segmentation are the two essential examples of this. The algorithms need to be reasonably fast, reliable, accurate, and highly automated. Information theory provides a means to create such systems. In this paper we present thresholding segmentation using image entropy and a registration technique based on maximization of mutual information. Then we show some experimental results using real-world computed tomography (CT) and medical resonance imaging (MRI) data.
The way brain networks maintain high transmission efficiency is believed to be fundamental in understanding brain activity. Brains consisting of more cells render information transmission more reliable and robust to noise. On the other hand, processing information in larger networks requires additional energy. Recent studies suggest that it is complexity, connectivity, and function diversity, rather than just size and the number of neurons, that could favour the evolution of memory, learning, and higher cognition. In this paper, we use Shannon information theory to address transmission efficiency quantitatively. We describe neural networks as communication channels, and then we measure information as mutual information between stimuli and network responses. We employ a probabilistic neuron model based on the approach proposed by Levy and Baxter, which comprises essential qualitative information transfer mechanisms. In this paper, we overview and discuss our previous quantitative results regarding brain-inspired networks, addressing their qualitative consequences in the context of broader literature. It is shown that mutual information is often maximized in a very noisy environment e.g., where only one-third of all input spikes are allowed to pass through noisy synapses and farther into the network. Moreover, we show that inhibitory connections as well as properly displaced long-range connections often significantly improve transmission efficiency. A deep understanding of brain processes in terms of advanced mathematical science plays an important role in the explanation of the nature of brain efficiency. Our results confirm that basic brain components that appear during the evolution process arise to optimise transmission performance.
5
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Information transmission over communication channels has been characterized by weighted information schemes involving probabilities and weights. Binary erasure channel has been used as an example for determination of constants in the proposed measure.
Registration is one of the essential medical image processing techniques. The goal is to find a geometric transformation, that relates corresponding voxels in two different 3D images of the same object. The publication presents a registration technique based on maximization of mutual information.
In this paper we present the objective video quality metric based on mutual information and Human Visual System. The calculation of proposed metric consists of two stages. In the first stage of quality evaluation whole original and test sequence are pre-processed by the Human Visual System. In the second stage we calculate mutual information which has been utilized as the quality evaluation criteria. The mutual information was calculated between the frame from original sequence and the corresponding frame from test sequence. For this testing purpose we choose Foreman video at CIF resolution. To prove reliability of our metric were compared it with some commonly used objective methods for measuring the video quality. The results show that presented objective video quality metric based on mutual information and Human Visual System provides relevant results in comparison with results of other objective methods so it is suitable candidate for measuring the video quality.
Perception is a constructive mental process, which cannot be considered impersonally. Similarly, music cannot be cognised solely on the basis of its score, since its coming into being is strictly connected to the activation of human memory and sound imagination. The patterns that emerge from the sounds of heard music enable the listener to draw conclusions regarding the structures those sounds embody. However, such conclusions are accompanied by a degree of uncertainty, which concerns not just the perceived moment of the heard music, but also the way in which it is represented in the listener’s memory. Perception is an inferential, multi-layered, uncertain process, in which particular patterns seem more likely than others. Mental representations of those probabilities lie behind such essential musical phenomena as surprise, tension, expectation and pitch identification, which are fixed elements of theperception of music. The aim of the present article is to describe the essence of three selected types of music modelling, based on spectral anticipation (Shlomo Dubnov), based on memory (Rens Bod), and exploiting the dynamic character of music to obtain information (Samer Abdallah and Mark Plumbley). All these models take account of the element of uncertainty that accompanies the perception of music; hence they make use the foundations of information theory and statistical analysis as measurement ‘tools’. The use of these tools makes it possible to obtain numerical rates, which inform us of the degree of predictability of the musical structures being analysed. One crucial advantage of these methods is the possibility of evaluating them in respect to the use of real musical structures, deriving from actual music, and not abstract structures formed for the purposes of research. We obtain cognitive insight into the analysed music by employing methods of a mathematical provenance, and so we have the possibility of examining music whilst taking account of the role of the listener, but with the use of objectivised methods.
In this paper the possibility of tracking the evolution of reconstructed pseudo-phase portraits in the diagnosis of positive-displacement pump wear has been presented. The reconstructed pseudo-phase portraits were obtained from vibration signals measured in characteristic places on the pump casing and from the dynamic pressure graphs recorded in the output port of the pump during the passive test experiment. The recorded measurement concerned to tree state of pump condition: in full working order, in part working order and pump with wear out elements.
PL
W artykule przedstawiono możliwość wykorzystania odtworzonych z zarejestrowanych sygnałów przebiegów pseudoportretów fazowych w diagnostyce zużycia pomp wyporowych na przykładzie badań pompy wielotłoczkowej. Odtworzone pseudoportrety fazowe otrzymano w wyniku przeprowadzenia biernego eksperymentu diagnostycznego, ze zmierzonych w charakterystycznych miejscach korpusu pompy sygnałów wibracji oraz dodatkowo z przebiegów ciśnienia dynamicznego dla trzech stanów pracy pompy: pełnej sprawności, stanu częściowej sprawności oraz pracy pompy zużytej.
We present the design of a platform for acquisition and digital processing of biosignals. The objective of this platform is to process biosignals in real-time to obtain quantitative indicators for joint analysis of biosignals ensembles. An important indicator of non-linear dependence between signals is the mutual information. The estimation of the mutual information between signals is time- and resource-consuming when using standard software implementations on normal computers. To circumvent the calculation limitations on standard software implementations we use a reconfigurable computing unit of type FPGA, were the calculation of mutual information is specified in hardware.
PL
Przedstawiamy projekt platformy służącej do pozyskiwania i cyfrowej obróbki biosygnałów. Jej zadaniem jest przetwarzanie biosygnałów w czasie rzeczywistym w celu uzyskania wskaźników ilościowych dla zintegrowanej analizy zespółów biosygnałów. Ważnym wskaźnikiem nieliniowej zależności pomiędzy sygnałami jest informacja wzajemna. Jej oszacowanie pomiędzy sygnałami przy użyciu standardowego oprogramowania na zwykłych komputerach jest mało wydajne i czasochłonne. Aby obejść ograniczenia narzucone przez narzędzia zwykle wykorzystywane w tym celu, zastosowano rekonfigurowalną jednostkę typu FPGA, w której obliczenia informacji wzajemnej są określone.
A presence of a noise is typical for real-world data. In order to avoid its negative impact on methods of time series analysis, noise reduction procedures may be used. The achieved results of an application of such procedures in identification of chaos or nonlinearity seem to be encouraging. One of the noise reduction methods is the Schreiber method, which, as it has been shown, is able to effectively reduce a noise added to time series generated by deterministic systems with chaotic dynamics. However, while analyzing real-world data, a researcher usually cannot be sure if the generating system is deterministic. Therefore, there is a risk that a noise reduction method will be applied to random data. In this paper, it has been shown that in situations where there in no clear evidence that investigated data are generated by a deterministic system, the Schreiber noise reduction method may negatively affect identification of time series. In the simulation carried out in this paper, the BDS test, the mutual information measure and the Pearson autocorrelation coefficient were used. The research has shown that an application of the Schreiber method may introduce spurious nonlinear dependencies to investigated data. As a result, random series may be misidentified as nonlinear.
PL
Jednym ze sposobów ograniczenia negatywnego wpływu obecności szumu losowego na analizę rzeczywistych szeregów czasowych jest stosowanie metod redukcji szumu. Prezentowane w literaturze przedmiotu rezultaty zastosowania takich procedur w procesie identyfikacji nieliniowości i chaosu są zachęcające. Jedną z metod redukcji szumu jest metoda Schreibera, która, jak wykazano, prowadzi do efektywnej redukcji szumu losowego dodanego do danych wygenerowanych z systemów deterministycznych o dynamice chaotycznej. Jednakże w przypadku danych rzeczywistych, badacz zwykle pozbawiony jest wiedzy, czy system generujący rzeczywiście jest deterministyczny. Istnieje więc ryzyko, że redukcji szumu zostaną wówczas poddane dane losowe. W niniejszym artykule wykazano, iż w sytuacji, gdy brak jest wyraźnych podstaw do stwierdzenia, że badany szereg pochodzi z systemu deterministycznego, metodę Schreibera należy stosować z dużą ostrożnością. Z przeprowadzonych symulacji, w których wykorzystano test BDS, miarę informacji wzajemnej oraz współczynnik korelacji liniowej Pearsona wynika bowiem, że redukcja szumu może wprowadzić do analizowanych danych, zależności o charakterze nieliniowym. W efekcie szeregi losowe mogą zostać błędnie zidentyfikowane jako nieliniowe.
In this paper we discuss in detail the resonance and oversampling features of the 0/1 test for chaos in continuous systems and propose methods to avoid those undesired features. Our method is based on certain frequency properties of the 0/1 test. When reconstructing the phase space, our approach is compared with the first minimum of the mutual information method. Several numerical results for typical chaotic systems (including memristive circuits) are included.
14
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Electroencephalogram (EEG) is the brain signal that contains the valuable information about different states of the brain. In this study EEG signals are analyzed for evaluating epileptic seizures in these signals and their sub-bands and comparing epileptic states with other states. A discrete wavelet transform is applied for decompose the EEGs into its sub-bands. The chaotic behavior of EEGs is evaluated by means ol normalized Shannon and spectral entropies. Entropy method is presented for detection of epileptic seizures through the analysis of EEGs and their sub-bands. At the end the mixture K-nearest neighbor and mutual information method is applied as a classifier to classify the different states in EEGs and their sub-bands. This method is applied to three different groups of EEG signals: 1) healthy states, 2) epileptic states during a seizure-free interval (interictal EEG), 3) epileptic states during a seizure (ictal EEG). The proposed method could classify different states with 99% accuracy.
PL
Elektroencefalografia EEG jest analizą sygnału mózgu. W artykule przedstawiono metody analizy sygnału EEG stosowane w celu wykrycia epilepsji. Zastosowano dyskretną transformatę falkową do dekompozycji sygnału EEG. Wykorzystano metodę entropii do detekcji sygnału związanego z epilepsją. Metody zastosowano do trzech grup pacjentów: zdrowych, chorych na epilepsję i chorych w czasie ataku epilepsji.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.