СЪСТОЯНИЕТО КАТО ПРИЧИНА: КОНСТРУКЦИИТЕ С КАКЪВТО И КАКТО С ПРИЧИННО ЗНАЧЕНИЕ / STATE AS A CAUSE: CAUSAL CONSTRUCTIONS WITH КАКЪВТО AND КАКТО

Apart from their main function, the Bulgarian constructions with какъвто and както can express causal relations between the situations described in the main clause and the subordinate clause. The paper shows that in the causal use of both constructions the subordinate clause expresses a static feature, serving as a justification of the situation presented in the main clause. In addition, there is a semantic differentiation between the two models: какъвто normally expresses a usual (permanent) feature, while както denotes mostly a temporary (episodic) feature. Observations on the means used by Russian translators to convey this kind of causal meaning are also put forward. Keywords: state, causal relations, relative clauses, manner clauses, Bulgarian, Russian

Reconfigurable Filter Design

This chapter discusses recent development of reconfigurable filters. The technical terminology reconfigurable means that a circuit is designed in a way to have various electrical characteristics comparing with one which has a static feature. For the filter design, the various electrical characteristics can be considered as the filter can tune its operating frequency, bandwidth, and/or have multiple operational modes, that is, bandstop or bandpass modes. Also, recently, the filters that can exhibit an improved impedance matching performance over its stopband have been reported. It provides more options for the filter designers to realize the reconfigurable filters having reflective and/or absorptive frequency response types to satisfy a prior given requirement. In this chapter, recently devised filter designs will be covered and essential frequency tuning elements to realize the reconfigurable characteristic will be introduced as well.

A Novel Approach to Learning Models on EEG Data Using Graph Theory Features - A Comparative Study

Functional Connectivity analysis using Electroencephalography signals is common. The EEG signals are converted to networks by transforming the signals into a correlation matrix and analyzing the resulting networks. Here, four learning models, namely, Logistic Regression, Random Forest, Support Vector Machine, and Recurrent Neural Networks, are implemented on the correlation matrix data to classify them either on their psychometric assessment or the effect of therapy; The EEG data is trail-based/event-related. The classifications based on RNN provided higher accuracy( 74-88%) than the other three models( 50-78%). Instead of using individual graph features, a correlation matrix provides an initial test of the data. When compared with the time-resolved correlation matrix, it offered a 4-5% higher accuracy. The time-resolved correlation matrix is better suited for dynamic studies here; it provides lower accuracy when compared to the correlation matrix, a static feature.

Dynamic Regularization on Activation Sparsity for Neural Network Efficiency Improvement

When deploying deep neural networks in embedded systems, it is crucial to decrease the model size and computational complexity for improving the execution speed and efficiency. In addition to conventional compression techniques, e.g., weight pruning and quantization, removing unimportant activations can also dramatically reduce the amount of data communication and the computation cost. Unlike weight parameters, the pattern of activations is directly related to input data and thereby changes dynamically. To regulate the dynamic activation sparsity (DAS), in this work, we propose a generic low-cost approach based on winners-take-all (WTA) dropout technique. The network enhanced by the proposed WTA dropout, namely DASNet , features structured activation sparsity with an improved sparsity level. Compared to the static feature map pruning methods, DASNets provide better computation cost reduction. The WTA dropout technique can be easily applied in deep neural networks without incurring additional training variables. More importantly, DASNet can be seamlessly integrated with other compression techniques, such as weight pruning and quantization, without compromising accuracy. Our experiments on various networks and datasets present significant runtime speedups with negligible accuracy losses.

Foreground Scattering Elimination by Inverse Lock-in-Like Spatial Modulation

We describe a simple approach to enhance vision, which is impaired by close range obscuring and/or scattering structures. Such structures may be found on a dirty windscreen of a car, or by tree branches blocking the vision of objects behind. The main idea is to spatially modulate the obscuration, either by periodically moving the detector/eye or by letting the obscuration modulate itself, such as branches swinging in the wind. The approach has similarities to electronic lock-in techniques, where the feature of interest is modulated to enable it to be isolated from the strong perturbing background, but now, we modulate the background instead to isolate the static feature of interest. Thus, the approach can be denoted as “inverse lock-in-like spatial modulation”. We also apply a new digital imaging processing technique based on a combination of the Interframe Difference and Gaussian Mixture models for digital separation between the objects of interest and the background, and make connections to the Gestalt vision psychology field.