Browsing by Subject "Energy Efficiency"

Now showing 1 - 3 of 3

Open Access
Covolutional neural networks based on non-euclidean operators
(2018-01) Badawi, Diaa Hisham Jamil
Dot product-based operations in neural net feedforwarding passes are replaced with an ℓ₁-norm inducing operator, which itself is multiplication-free. The neural net, which is called AddNet, retains attributes of ℓ₁-norm based feature extraction schemes such as resilience against outliers. Furthermore, feedforwarding passes can be realized using fewer multiplication operations, which implies energy efficiency. The ℓ₁-norm inducing operator is differentiable w.r.t its operands almost everywhere. Therefore, it is possible to use it in neural nets that are to be trained through standard backpropagation algorithm. AddNet requires scaling (multiplicative) bias so that cost gradients do not explode during training. We present different choices for multiplicative bias: trainable, directly dependent upon the associated weights, or fixed. We also present a sparse variant of that operator, where partial or full binarization of weights is achievable. We ran our experiments over MNIST and CIFAR-10 datasets. AddNet could achieve results that are 0:1% less accurate than a ordinary CNN. Furthermore, trainable multiplicative bias helps the network to converge fast. In comparison with other binary-weights neural nets, AddNet achieves better results even with full or almost full weight magnitude pruning while keeping the sign information after training. As for experimenting on CIFAR-10, AddNet achieves accuracy 5% less than a ordinary CNN. Nevertheless, AddNet is more rigorous against impulsive noise data corruption and it outperforms the corresponding ordinary CNN in the presence of impulsive noise, even at small levels of noise.
Open Access
Performance analysis of the carrier-sense multiple access protocol for future generation wireless networks
(2013) Köseoğlu, Mehmet
Variants of the carrier-sense multiple access (CSMA) protocol has been employed in many communications protocols such as the IEEE 802.11 and Ethernet standards. CSMA based medium access control (MAC) mechanisms have been recently proposed for other communications scenarios such as sensor networks and acoustical underwater networks. Despite its widespread use, the performance of the CSMA protocol is not well-studied from the perspective of these newly encountered networking scenarios. We here investigate the performance of the CSMA protocol from the point of three different aspects: throughput in networks with large propagation delay, short-term fairness for delay sensitive applications in large networks and energy efficiency-throughput trade-off in networks with battery operated devices. Firstly, we investigate the performance of the CSMA protocol for channels with large propagation delay. Such channels are recently encountered in underwater acoustic networks and in terrestrial wireless networks covering larger areas. However, a mathematical model of CSMA performance in such networks is not known. We propose a semi-Markov model for a 2-node CSMA channel and then extend this model for arbitrary number of users. Using this model, we obtain the optimum symmetric probing rate that achieves the maximum network throughput as a function of the average propagation delay, ¯d, and the number of nodes sharing the channel, N. The proposed model predicts that the total capacity decreases with ¯d −1 as N goes to infinity when all nodes probe the channel at the optimum rate. The optimum probing rate for each node decreases with 1/N and the total optimum probing rate decreases faster than ¯d −1 as N goes to infinity. Secondly, we investigate whether the short-term fairness of a large CSMA network degrades with the network size and density. Our results suggest that (a) the throughput region that can be achieved within the acceptable limits of shortterm fairness reduces as the number of contending neighboring nodes increases for random regular conflict graphs, (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 topology. 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. Thirdly, we investigate the energy efficiency of a CSMA network proposing a model for the energy consumption of a node as a function of its throughput. We show that operating the CSMA network at a very high or at a very low throughput is energy inefficient because of increasing carrier-sensing and sleeping costs, respectively. Achieving a balance between these two opposite operating regimes, we derive the energy-optimum carrier-sensing rate and the energy-optimum throughput which maximize the number of transmitted bits for a given energy budget. For the single-hop case, we show that the energy-optimum total throughput increases as the number of nodes sharing the channel increases. For the multi-hop case, we show that the energy-optimum throughput decreases as the degree of the conflict graph of the network increases. For both cases, the energy-optimum throughput reduces as the power required for carrier-sensing increases. The energy-optimum throughput is also shown to be substantially lower than the maximum throughput and the gap increases as the degree of the conflict graph increases for multi-hop networks.
Open Access
Real-time routing with priority scheduling and power adjustment in wireless sensor networks
(2008) Çelikkaya, Emine Büşra
Many wireless sensor network applications require real-time communication, and real-time applications require packets to reach destination on time. However, applications may send packets with different priorities and hence delay bounds for packets may vary significantly. Therefore packet differentiation in the network is essential for meeting the deadline requirements. We propose a routing protocol that supports real-time communication by utilizing transmit power adjustment in order to meet the deadline of urgent packets and use energy efficiently. Our protocol also provides packet scheduling and gives precedence to urgent packets. We have conducted experiments on our sensor network testbed to observe the effects of transmit power on end-to-end delay. As expected, increasing transmit power increases the range and link quality, and reduces the number of hops to reach destination. Therefore adjusting transmit power has a great effect on delivery time and can reduce the end-to-end delay. Our protocol, Real-time Routing with Priority Scheduling and Power Adjustment, uses different levels of transmit power for packets with different priorities. It sends urgent packets with maximum power to minimize end-to-end delay and lower priority packets with reduced power to save energy and balance the load on nodes. Simulation results show that our routing protocol increases the deadline meet ratio of packets and reduces the transmit energy spent per packet when compared to routing protocols that use fixed transmit power. Additionally, results indicate that our approach lessens the interference on sensor nodes that are caused by other transmissions and helps balancing the load on the nodes.