Browsing by Subject "Efficient"

Now showing 1 - 4 of 4

Open Access
Effect of Network Density and Size on the Short-Term Fairness Performance of CSMA Systems
(Springer Open Journal, 2012) Koseoglu, M.; Karasan, E.; Alanyali, M.
As the penetration of wireless networks increase, number of neighboring networks contending for the limited unlicensed spectrum band increases. This interference between neighboring networks leads to large systems of locally interacting networks. We investigate whether the short-term fairness of this system of networks degrades with the system size and density if transmitters employ random spectrum access with carrier sensing (CSMA). Our results suggest that (a) short-term fair capacity, which is the throughput region that can be achieved within the acceptable limits of short-term fairness, reduces as the number of contending neighboring networks, i.e., degree of the conflict graph, increases for random regular conflict graphs where each vertex has the same number of neighbors, (b) short-term fair capacity weakly depends on the network size for a random regular conflict graph but a stronger dependence is observed for a grid deployment. We demonstrate the implications of this study on a city-wide Wi-Fi network deployment scenario by relating the short-term fairness to the density of deployment. We also present related results from the statistical physics literature on long-range correlations in large systems and point out the relation between these results and short-term fairness of CSMA systems.
Embargo
Efficient neural network processing via model compression and low-power functional units
(2024-12) Karakuloğlu, Ali Necat
We present a framework that contributes neural network optimization through novel methods in pruning, quantization, and arithmetic unit design for resource-constrained devices to datacenters. The first component is a pruning method that employs an importance metric to measure and selectively eliminate less critical neurons and weights, achieving high compression rates up to 99.9% without sacrificing significant accuracy. This idea is improved by a novel pruning schedule that optimizes the balance between compression and model’s generalization capa-bility. Next, we introduce a quantization method that combines with pruning to improve hardware compatibility for floating point format, offering efficient model compression and fast computation and general usability. Finally, we propose a logarithmic arithmetic unit that designed as an energy-efficient alternative to conventional floating-point operations, providing precise and configurable processing without relying on bulky lookup tables. Extensive evaluations across different datasets and CUDA-based simulations and Verilog based hardware designs indicate that our approaches outperforms existing methods, making it a powerful solution for deploying artificial intelligence models more efficiently.
Open Access
Efficient NP tests for anomaly detection over birth-death type DTMCs
(Springer New York LLC, 2018) Özkan, H.; Özkan, F.; Delibalta, I.; Kozat, Süleyman S.
We propose computationally highly efficient Neyman-Pearson (NP) tests for anomaly detection over birth-death type discrete time Markov chains. Instead of relying on extensive Monte Carlo simulations (as in the case of the baseline NP), we directly approximate the log-likelihood density to match the desired false alarm rate; and therefore obtain our efficient implementations. The proposed algorithms are appropriate for processing large scale data in online applications with real time false alarm rate controllability. Since we do not require parameter tuning, our algorithms are also adaptive to non-stationarity in the data source. In our experiments, the proposed tests demonstrate superior detection power compared to the baseline NP while nearly achieving the desired rates with negligible computational resources.
Open Access
Online anomaly detection under Markov statistics with controllable type-I error
(Institute of Electrical and Electronics Engineers Inc., 2016) Ozkan, H.; Ozkan, F.; Kozat, S. S.
We study anomaly detection for fast streaming temporal data with real time Type-I error, i.e., false alarm rate, controllability; and propose a computationally highly efficient online algorithm, which closely achieves a specified false alarm rate while maximizing the detection power. Regardless of whether the source is stationary or nonstationary, the proposed algorithm sequentially receives a time series and learns the nominal attributes - in the online setting - under possibly varying Markov statistics. Then, an anomaly is declared at a time instance, if the observations are statistically sufficiently deviant. Moreover, the proposed algorithm is remarkably versatile since it does not require parameter tuning to match the desired rates even in the case of strong nonstationarity. The presented study is the first to provide the online implementation of Neyman-Pearson (NP) characterization for the problem such that the NP optimality, i.e., maximum detection power at a specified false alarm rate, is nearly achieved in a truly online manner. In this regard, the proposed algorithm is highly novel and appropriate especially for the applications requiring sequential data processing at large scales/high rates due to its parameter-tuning free computational efficient design with the practical NP constraints under stationary or non-stationary source statistics. © 2015 IEEE.