arXiv subjects

Ivana Nikoloska

Publications and source records attributed to Ivana Nikoloska.

25 records · Page 2Linked to original sources

arXiv2022

Training Hybrid Classical-Quantum Classifiers via Stochastic Variational Optimization

Quantum machine learning has emerged as a potential practical application of near-term quantum devices. In this work, we study a two-layer hybrid classical-quantum classifier in which a first layer of quantum stochastic neurons implementing generalized linear models (QGLMs) is followed by a second classical combining layer. The input to the first, hidden, layer is obtained via amplitude encoding in order to leverage the exponential size of the fan-in of the quantum neurons in the number of qubits per neuron. To facilitate implementation of the QGLMs, all weights and activations are binary. While the state of the art on training strategies for this class of models is limited to exhaustive search and single-neuron perceptron-like bit-flip strategies, this letter introduces a stochastic variational optimization approach that enables the joint training of quantum and classical layers via stochastic gradient descent. Experiments show the advantages of the approach for a variety of activation functions implemented by QGLM neurons.

quant-ph↗

arXiv2021

Fast Power Control Adaptation via Meta-Learning for Random Edge Graph Neural Networks

Power control in decentralized wireless networks poses a complex stochastic optimization problem when formulated as the maximization of the average sum rate for arbitrary interference graphs. Recent work has introduced data-driven design methods that leverage graph neural network (GNN) to efficiently parametrize the power control policy mapping channel state information (CSI) to the power vector. The specific GNN architecture, known as random edge GNN (REGNN), defines a non-linear graph convolutional architecture whose spatial weights are tied to the channel coefficients, enabling a direct adaption to channel conditions. This paper studies the higher-level problem of enabling fast adaption of the power control policy to time-varying topologies. To this end, we apply first-order meta-learning on data from multiple topologies with the aim of optimizing for a few-shot adaptation to new network configurations.

cs.IT↗

arXiv2020

Inference over Wireless IoT Links with Importance-Filtered Updates

We consider a communication cell comprised of Internet-of-Things (IoT) nodes transmitting to a common Access Point (AP). The nodes in the cell are assumed to generate data samples periodically, which are to be transmitted to the AP. The AP hosts a machine learning model, such as a neural network, which is trained on the received data samples to make accurate inferences. We address the following tradeoff: The more often the IoT nodes transmit, the higher the accuracy of the inference made by the AP, but also the higher the energy expenditure at the IoT nodes. We propose a data filtering scheme employed by the IoT nodes, which we refer to as distributed importance filtering in order to filter out redundant data samples already at the IoT nodes. The IoT nodes do not have large on-device machine learning models and the data filtering scheme operates under periodic instructions from the model placed at the AP. The proposed scheme is evaluated using neural networks on a benchmark machine vision dataset, as well as in two practical scenarios: leakage detection in water distribution networks and air-pollution detection in urban areas. The results show that the proposed scheme offers significant benefits in terms of network longevity as it preserves the devices' resources, whilst maintaining high inference accuracy. Our approach reduces the the computational complexity for training the model and obviates the need for data pre-processing, which makes it highly applicable in practical IoT scenarios.

eess.SP↗

arXiv2020

Deep Reinforcement Learning-Aided Random Access

We consider a system model comprised of an access point (AP) and K Internet of Things (IoT) nodes that sporadically become active in order to send data to the AP. The AP is assumed to have N time-frequency resource blocks that it can allocate to the IoT nodes that wish to send data, where N < K. The main problem is how to allocate the N time-frequency resource blocks to the IoT nodes in each time slot such that the average packet rate is maximized. For this problem, we propose a deep reinforcement learning (DRL)-aided random access (RA) scheme, where an intelligent DRL agent at the AP learns to predict the activity of the IoT nodes in each time slot and grants time-frequency resource blocks to the IoT nodes predicted as active. Next, the IoT nodes that are missclassified as non-active by the DRL agent, as well as unseen or newly arrived nodes in the cell, employ the standard RA scheme in order to obtain time-frequency resource blocks. We leverage expert knowledge for faster training of the DRL agent. Our numerical results show significant improvements in terms of average packet rate when the proposed DRL-aided RA scheme is implemented compared to the existing solution used in practice, the standard RA scheme.

cs.IT↗

arXiv2019

On the Secrecy Capacity of a Full-Duplex Wirelessly Powered Communication System

In this paper, we investigate the secrecy capacity of a point-to-point, full-duplex (FD) wirelesly powered communication system in the presence of a passive eavesdropper (EVE). The considered system is comprised of an energy transmitter (ET), an energy harvesting user (EHU), and a passive EVE. The ET transmits radio-frequency energy, which is used for powering the EHU as well as for generating interference at the EVE. The EHU uses the energy harvested from the ET to transmit confidential messages back to the ET. As a consequence of the FD mode of operation, both the EHU and the ET are subjected to self-interference, which has different effects at the two nodes. In particular, the self-interference impairs the decoding of the received message at the ET, whilst it serves as an additional energy source at the EHU. For this system model, we derive an upper and a lower bound on the secrecy capacity. For the lower bound, we propose a simple achievability scheme. Our numerical results show significant improvements in terms of achievable secrecy rate when the proposed communication scheme is employed against its half-duplex counterpart, even for practical self-interference values at the ET.

cs.IT↗

arXiv2017

Capacity of a Full-Duplex Wirelessly Powered Communication System with Self-Interference and Processing Cost

In this paper, we investigate the capacity of a point-to-point, full-duplex (FD), wirelessly powered communication system impaired by self-interference. This system is comprised of an energy transmitter (ET) and an energy harvesting user (EHU), both operating in a FD mode. The ET transmits energy towards the EHU. The EHU harvests this energy and uses it to transmit information back to the ET. As a result of the FD mode, both nodes are affected by self-interference. The self-interference has a different effect at the two nodes: it impairs the decoding of the received signal at the ET, however, it provides an additional source of energy for the EHU. This paper derives the capacity of this communication system assuming a processing cost at the EHU and additive white Gaussian noise channel with block fading. Thereby, we show that the capacity achieving scheme is relatively simple and therefore applicable to devices with limited resources. Moreover, our numerical results show significant improvements in terms of data rate when the capacity achieving strategy is employed compared to half-duplex transmission. Moreover, we show the positive and negative effects of the self-interference at the EHU and the ET, respectively. Furthermore, we show the crippling effect of the processing cost and demonstrate that failing to take it into consideration gives a false impression in terms of achievable rate.

cs.IT↗

arXiv2016

Wireless Networks with Energy Harvesting and Power Transfer: Joint Power and Time Allocation

In this paper, we consider wireless powered communication networks which could operate perpetually, as the base station (BS) broadcasts energy to the multiple energy harvesting (EH) information transmitters. These employ "harvest then transmit" mechanism, as they spend all of their energy harvested during the previous BS energy broadcast to transmit the information towards the BS. Assuming time division multiple access (TDMA), we propose a novel transmission scheme for jointly optimal allocation of the BS broadcasting power and time sharing among the wireless nodes, which maximizes the overall network throughput, under the constraint of average transmit power and maximum transmit power at the BS. The proposed scheme significantly outperforms "state of the art" schemes that employ only the optimal time allocation. If a single EH transmitter is considered, we generalize the optimal solutions for the case of fixed circuit power consumption, which refers to a much more practical scenario.

cs.IT↗