SUNY Polytechnic Institute

We investigate the gravitational wave background predicted by a two-scalar-field cosmological model that aims to unify primordial inflation with the dark sector, namely, late-time dark energy and dark matter, in a single and self-consistent theoretical framework. The model is constructed from an action inspired by several extensions of general relativity and string-inspired scenarios and features a nonminimal interaction between the two scalar fields, while both remain minimally coupled to gravity. In this context, we derive the gravitational wave energy spectrum over wavelengths ranging from today’s Hubble horizon to those at the end of inflation. We employ the continuous Bogoliubov coefficient formalism, originally introduced to describe particle creation in an expanding Universe, in analogy to the well-established mechanism of gravitational particle production and, in particular, generalized to gravitons. Using this method, which enables an accurate description of graviton creation across all cosmological epochs, we find that inflation provides the dominant gravitational wave contribution, while subdominant features arise at the inflation-radiation, radiation-matter, and matter-dark energy transitions, i.e., epochs naturally encoded inside our scalar field picture. The resulting energy density spectrum is thus compared with the sensitivity curves of the planned next-generation ground- and space-based gravitational wave observatories. The comparison identifies frequency bands where the predicted signal could be probed, providing those windows associated with potentially detectable signals, bounded by our analyses. Consequences of our recipe are thus compared with numerical outcomes and the corresponding physical properties are discussed in detail.

This paper explores the use of Machine Learning (ML) techniques in enhancing malware detection amidst the evolving landscape of cyber threats. Traditional signature-based methods are becoming increasingly ineffective against sophisticated malware employing evasion tactics like polymorphism and metamorphism. Our study introduces a taxonomy of ML-based malware detection models, including supervised, unsupervised, and reinforcement learning approaches. We conduct an evaluation of these models on various datasets to determine their effectiveness in identifying malicious code. The paper discusses the strengths, weaknesses, and security concerns of these ML models, particularly their vulnerability to adversarial attacks. Among the techniques evaluated, basic supervised learning algorithms, such as Decision Trees and Random Forests, emerged as the most effective. These algorithms demonstrated superior performance in detecting known malware types, characterized by high accuracy, ease of implementation, and efficient processing times compared to more complex models. Our findings offer insights for developing more resilient and intelligent malware detection systems and address the challenges of integrating these models into real-world cybersecurity systems.

Artificial-intelligence-powered edge devices are inspiring interest in always-on, intelligent, and self-powered visual perception systems. Due to the high energy cost of converting raw data and the limited computing and energy resources available, designing energy-efficient and low bandwidth CMOS vision sensors is vital as these emerging systems require continuous sensing and instant processing. This paper proposes a low-power integrated sensing and computing engine, namely <monospace>APRIS</monospace>, including a novel software/hardware co-design technique. This method provides a highly parallel analog multiplication and accumulation-in-pixel scheme, which realizes low-precision quantized weight neural networks to mitigate the overhead of analog-to-digital converters and analog buffers. Moreover, in order to reduce the size and power consumption, we propose the implementation of an approximate ADC in the readout circuit. Our system utilizes eight memory banks to increase computation parallelism, which has a dramatic effect on its speed and efficiency. Moreover, the proposed structure supports a zero-skipping scheme to reduce power consumption further. Our circuit-to-application co-simulation results demonstrate a comparable accuracy for our platform to the full-precision baseline on various object classification tasks while reaching an efficiency of <inline-formula><tex-math notation="LaTeX">$\sim$</tex-math></inline-formula>3.48 TOp/s/W.

This study quantitatively evaluates the throughput loss of a non-terrestrial network (NTN) user equipment (UE) as a function of its isolation distance from a terrestrial network (TN) cluster when the TN and NTN operate in adjacent frequency bands. A 5 % throughput loss was defined as the interference threshold, and the minimum isolation distance required to satisfy this criterion was derived based on both the average and the 5th-percentile throughput loss values. The results indicate that the minimum isolation distance is approximately 0.197 km for the average case and 1.036 km for the 5th-percentile case, corresponding to a relaxation of about 1.3 km and 0.5 km, respectively, compared with the 1.5 km isolation distance recommended by 3GPP.

This paper presents a novel design and analysis of a wideband metasurface-enhanced antenna aimed at polarization conversion and radar cross-section (RCS) reduction. The proposed antenna combines an octagonal patch structure with a metasurface made of periodic fishbone unit cells, providing improved polarization conversion efficiency and enhanced control over wave propagation. Designed to operate in the 2030 GHz frequency range, the antenna achieves a notable RCS reduction of up to 20 dB and maintains strong polarization conversion performance across a wide range of incidence angles. Simulation results confirm the design's ability to suppress backscatter and improve both co-polarized and crosspolarized reflection characteristics, making it well-suited for stealth and modern communication applications.