Timur Düzenli

Bu çalışmada, otomatik modülasyon sınıflandırması için ResNet-50 derin öğrenme mimarisi ile işaret grafiği ve ayrık dalgacık dönüşümü tabanlı özniteliklerin birleştirilmesi yaklaşımı önerilmiştir. Çalışmanın amacı, gürültülü haberleşme kanallarında modülasyon türlerinin sınıflandırma başarımını artırmak amacıyla sinyallerin hem zaman hem de frekans alanı özelliklerini birlikte kullanmaktır. RadioML 2018.01A veri setinden seçilen dört temel modülasyon türü (OOK, BPSK, QPSK ve 16QAM) üzerinde benzetimler gerçekleştirilmiştir. Her bir sinyal için Coiflet-2 (coif2) dalgacığı kullanılarak yüksek frekans bileşenlerini temsil eden detay katsayıları elde edilmiştir. Daha sonra, bu katsayılar ile gözlenen işaretin grafikleri birleştirilerek derin öğrenme modeline girdi olarak verilmiştir. Farklı sinyal gürültü oranı koşullarında yapılan benzetimlerde, önerilen modelin düşük sinyal-gürültü oranı değerlerinde dahi kararlı bir şekilde çalıştığı ve yüksek doğruluk oranlarına ulaştığı gözlenmiştir. Sinyal-gürültü oranı arttıkça doğruluk oranında beklendiği üzere yükselme gerçekleşmiş ve 30 dB seviyesinde %98’in üzerinde bir sınıflandırma başarısı elde edilmiştir. Bu sonuçlar, dalgacık tabanlı özniteliklerin modelin hem yapısal hem de geçici sinyal özelliklerini yakalama kabiliyetini güçlendirdiğini ve benzer modülasyon türleri arasındaki karışıklığı azalttığını göstermektedir.

In this paper, we investigate the geometric properties of a specific subclass of analytic functions satisfying the condition f' (z) ≺ cosh(√ z) meaning that the function f '(z) is subordinate to the function cosh(√ z). Also, we focus on deriving sharp inequalities for Taylor coefficients, particularly for b2 and the modulus of the second derivative f''(z). Utilizing the Schwarz lemma, both on the unit disc and on its boundary, we provide essential insights into the distortion and growth behaviors of these functions. The paper demonstrates the sharpness of these inequalities through extremal functions and applies the Julia–Wolff lemma to establish boundary behavior results. These findings contribute significantly to the understanding of the analytic functions associated with the hyperbolic cosine function, with potential applications in geometric function theory. It is considered that the extremal functions obtained in this study could be potential hyperbolic activation functions in neural network architectures. This perspective builds a conceptual bridge between geometric function theory and artificial intelligence, indicating that insights from complex analysis can inspire the development of more effective and theoretically grounded activation mechanisms in deep learning. Empirical evaluation of architectures built with novel activation functions may be considered as potential future work.

Underwater wireless optical communication systems face significant challenges due to the heterogeneous nature of the underwater environment and the attenuation of optical signals caused by absorption and scattering. These effects restrict the data transfer capacity and transmission distance, resulting in communication errors. Different modulation techniques are used to minimize the effects of these parameters. Automatic modulation classification plays a critical role in terms of effective management of spectrum resources. In this study, underwater wireless optical communication channels are modulated with different modulation techniques, and the signals are transformed into the discrete wavelet space, resulting in approximation and detail coefficients that are used as feature vectors for training machine learning algorithms. In addition, optimized classification features are determined for different signal-to-noise ratios and different transmission distances using the genetic algorithm. The results show that the approximation and detail coefficient energies provide higher classification performance in the classification of modulated signals according to statistical features such as mean, variance, and standard deviation. According to simulation results, an average classification accuracy of 82% has been obtained using the proposed discrete wavelet transform and genetic algorithm-based technique, which demonstrates high classification accuracy for noisy underwater channels.

In this brief, positive real functions are considered as driving point impedance functions, $Z(s)$ , which are utilized in electrical engineering for characteristic representation of circuits. Accordingly, for the real part of $Z(s)$ , sharp inequalities are presented with their extremal functions by using zeros and poles of $Z(s)$ . For these extremal functions, corresponding generic circuit schematics are given with their spectral analyses. According to the obtained results, it has been shown that various circuits can be obtained by using the extremal functions of the inequalities presented in this brief.

In this paper, an inequality for a transfer function is obtained assuming that its residues at the poles located on the imaginary axis in the right half plane. In addition, the extremal function of the proposed inequal-ity is obtained by performing sharpness analysis. To interpret the results of analyses in terms of control theory, root-locus curves are plotted. According to the results, marginally and asymptotically stable transfer functions can be determined using the obtained extremal function in the proposed theorem.

Driving point impedance functions (DPIFs) are frequently used in electrical engineering, and they represent characteristic properties of various types of circuits such as RL, RC, LC and RLC networks. In this paper, boundary analysis of driving point impedance functions are investigated using Schwarz lemma. Assuming that the driving point impedance function, Z(s), is given as Z(s)=A/2+c_p (s-1)^p+c_(p+1) (s-1)^(p+1)+... and it is analytic in the right half of the s-plane, novel boundaries are obtained for |Z^' (0)|. Accordingly, it is aimed to obtain novel inequalities which presents higher boundaries for |Z'(0)| and derive novel generic driving point impedace functions by performing extremal analysis of these obtained inequalities. It is also aimed to investigate how |Z'(s)| can be interpreted when it is considered at the boundary. According to simulation results, frequency characteristics of obtained driving point impedance functions can be used to design of multi-notch filters which are localized at certain frequency values.