Complexity Trade-off Research Articles

Hardware Implementation of a Softmax-Like Function for Deep Learning

In this paper a simplified hardware implementation of a CNN softmax-like layer is proposed. Initially the softmax activation function is analyzed in terms of required numerical accuracy and certain optimizations are proposed. A proposed adaptable hardware architecture is evaluated in terms of the introduced error due to the proposed softmax-like function. The proposed architecture can be adopted to the accuracy required by the application by retaining or eliminating certain terms of the approximation thus allowing to explore accuracy for complexity trade-offs. Furthermore, the proposed circuits are synthesized in a 90 nm 1.0 V CMOS standard-cell library using Synopsys Design Compiler. Comparisons reveal that significant reduction is achieved in area × delay and power × delay products for certain cases, respectively, over prior art. Area and power savings are achieved with respect to performance and accuracy.

Knowledge complexity and the mechanisms of knowledge generation and exploitation: The European evidence

A knowledge complexity trade-off can be identified if and when the complexity of the stock of knowledge engenders positive effects in the recombinant generation of new technological knowledge but negative ones in its exploitation in terms of productivity gains. On the one hand, the complexity of the stock of knowledge increases the scope for recombination and hence the amount of knowledge that each firm is able to generate with a given budget. On the other, the complexity of the stock of knowledge has controversial effects on productivity: the indirect effects -via the larger amount of knowledge generated upstream- are positive but the direct ones can be negative if they generate difficulties in the exploitation of a highly heterogeneous stock of knowledge. The econometric test on the European regions in the years 1997–2009, provides strong empirical evidence about the relevance of the composition of the stock of quasi-public knowledge and its twin positive effects in terms of the direct support of the knowledge generation function and of the indirect one on productivity growth. Moreover, the analysis highlights the negative direct effects on productivity dynamics, though with substantial heterogeneity across different groups of regions.

Atlasing of Assembly Landscapes using Distance Geometry and Graph Rigidity.

This Article describes a novel geometric methodology for analyzing free energy and kinetics of assembly driven by short-range pair-potentials in an implicit solvent and provides a proof-of-concept illustration of its unique capabilities. An atlas is a labeled partition of the assembly landscape into a roadmap of maximal, contiguous, nearly-equipotential-energy conformational regions or macrostates, together with their neighborhood relationships. The new methodology decouples the roadmap generation from sampling and produces: (1) a queryable atlas of local potential energy minima, their basin structure, energy barriers, and neighboring basins; (2) paths between a specified pair of basins, each path being a sequence of conformational regions or macrostates below a desired energy threshold; and (3) approximations of relative path lengths, basin volumes (configurational entropy), and path probabilities. Results demonstrating the core algorithm's capabilities and high computational efficiency have been generated by a resource-light, curated open source software implementation EASAL (Efficient Atlasing and Search of Assembly Landscapes, ACM Trans. Math. Softw. 2018 44, 1-48. 10.1145/3204472; see software, Efficient Atlasing and Search of Assembly Landscapes, 2016. https://bitbucket.org/geoplexity/easal; video, Video Illustrating the opensource software EASAL, 2016. https://cise.ufl.edu/~sitharam/EASALvideo.mpeg; and user guide, EASAL software user guide, 2016. https://bitbucket.org/geoplexity/easal/src/master/CompleteUserGuide.pdf). Running on a laptop with Intel(R) Core(TM) i7-7700@3.60 GHz CPU with 16GB of RAM, EASAL atlases several hundred thousand conformational regions or macrostates in minutes using a single compute core. Subsequent path and basin computations each take seconds. A parallelized EASAL version running on the same laptop with 4 cores gives a 3× speedup for atlas generation. The core algorithm's correctness, time complexity, and efficiency-accuracy trade-offs are formally guaranteed using modern distance geometry, geometric constraint systems and combinatorial rigidity. The methodology further links the shape of the input assembling units to a type of intuitive and queryable bar-code of the output atlas, which in turn determine stable assembled structures and kinetics. This succinct input-output relationship facilitates reverse analysis and control toward design. A novel feature that is crucial to both the high sampling efficiency and decoupling of roadmap generation from sampling is a recently developed theory of convex Cayley (distance-based) custom parametrizations specific to assembly, as opposed to folding. Representing microstates with macrostate-specific Cayley parameters, to generate microstate samples, avoids gradient-descent search used by all prevailing methods. Further, these parametrizations convexify conformational regions or macrostates. This ratchets up sampling efficiency, significantly reducing number of repeated and discarded samples. These features of the new stand-alone methodology can also be used to complement the strengths of prevailing methodologies including Molecular Dynamics, Monte Carlo, and Fast Fourier Transform based methods.

CMFALL: A Cascade and Parallel Multi-State Fall Detection Algorithm Using Waist-Mounted Tri-Axial Accelerometer Signals

As one of the main threats to people's health, especially for the elderly, falls have caused a large number of accidents. Detecting falls in time can minimize the severity of the injury and save lives. Therefore, designing fall detection algorithms to assist people's daily life has been a hot research topic. In the last decade, different fall detection approaches based on wearable sensors have been proposed. However, since the hardware resources of wearable sensors are very limited, designing accurate but energy-efficient fall detection algorithms remains an open challenge. To deal with this, an accurate but low-cost fall detection algorithm is proposed in this paper. Particularly, a novel cascade and parallel method that efficiently employs the characteristics of human falls and the advanced modeling ability of the neural network is proposed. Also, a novel design of a lightweight convolutional neural network with self-attention is proposed to achieve the best recognition/numerical complexity tradeoff. The proposed method, namely CMFALL, is evaluated together with a multitude of state-of-the-art models on a large dataset, where it performs the best with an F1-score exceeding 99%. Meanwhile, its computational cost and model size are only a few thousandths of other models. Such low computational cost and small size not only enable to embed it in a wearable sensor but also make the system power requirements quite low, which can enhance the autonomy of the wearable fall detector.

A Power-Efficient Spectrum-Sensing Scheme Using 1-Bit Quantizer and Modified Filter Banks

Spectrum sensing is an efficient way to determine the spectrum availabilities over the frequency range of interest, aiding in improving the spectrum utilization in the cognitive radio (CR) systems. Conventional Nyquist multiband sensing entails higher computational capability for sampling, quantization, and subsequent processing, lending the approach infeasible for applications with limited power budgets. In this brief, a power-efficient spectrum-sensing technique is proposed, which explores an accuracy–complexity tradeoff. The presented spectrum-sensing architecture is based on 1-bit quantization at the CR receiver and implements it in hardware by a resource- and power-efficient approach, using a finite-impulse-response (FIR) filter-bank channelizer. The proposed scheme allows the complex operators like multipliers and quantizers to be replaced by the inverter logic and high-speed comparators, reducing the hardware complexity and power consumption. We validate the proposed scheme on a field-programmable gate-array (FPGA) emulator for an aeronautical $L$ -band digital aeronautical communication system (LDACS) application, and our results show that the proposed scheme achieves substantial resource reduction with at most 5% degradation in the detection accuracy in this case.

Utility-Consistent Valuation Schemes for the Own Risk and Solvency Assessment of Life Insurance Companies

Abstract In this paper, we construct new valuation schemes for the liabilities and economic capital of insurance companies. Specifically, we first build a ‘SAHARA’ valuation framework based on Symmetric Asymptotic Hyperbolic Absolute Risk Aversion utility functions. Then, we construct a ‘SAHARA-CPT’ framework that incorporates the previous utility function as a value function and that is based on Cumulative Prospect Theory. The process used for assessing a life insurance company’s own funds consists in replacing the market-consistent parametrization with a utility-consistent parametrization that accounts for the risk aversion of the market and the long-term duration of the company’s commitments. Our illustrations show that this approach leads to a lower value of the Own Risk and Solvency Assessment and to a lower volatility of own funds. The framework that is based on cumulative prospect theory has the advantage over the expected utility theory framework that it considers a precautionary overweighting of extreme events, as a tradeoff for additional model complexity.

On the Accuracy–Complexity Tradeoff of Fuzzy Broad Learning System

The fuzzy broad learning system (FBLS) is a recently proposed neuro-fuzzy model that shares the similar structure of a broad learning system (BLS). It shows high accuracy in both classification and regression tasks and inherits the fast computational nature of a BLS. However, the ensemble of several fuzzy subsystems in an FBLS decreases the possibility of understanding the fuzzy model since the fuzzy rules from different fuzzy systems are difficult to combine together while keeping the consistence. To balance the model accuracy and complexity, a synthetically simplified FBLS with better interpretability, named compact FBLS (CFBLS), is developed in this article, which can generate much fewer and more explainable fuzzy rules for understanding. In such a way, only one traditional Takagi–Sugeno–Kang fuzzy system is employed in the feature layer of a CFBLS, and the input universe of discourse is equally partitioned to obtain the fuzzy sets with proper linguistic labels accordingly. The random feature selection matrix and rule combination matrix are employed to reduce the total number of fuzzy rules and to avoid the “curse of dimensionality.” The enhancement layer is kept in the CFBLS that helps to add a nonlinear transformation of input features to the traditional first-order polynomial used in the consequent part of a fuzzy rule. The pseudoinverse is also used to determine the parameters of CFBLS guaranteeing its fast computational nature. The experiments on the popular UCI and KEEL datasets indicate that the CFBLS can generate a smaller set of comprehensible fuzzy rules and achieve much higher accuracy than some state-of-the-art neuro-fuzzy models. Moreover, the advantage of CFBLS is also verified in a real-world application.

Machine Learning Aided Blockchain Assisted Framework for Wireless Networks

Inspired by its success in financial sectors, the blockchain technique is emerging as an enabling technology for secure distributed control and management of wireless networks. In order to fully benefit from this distributed ledger technology, its limitations, cost, complexity and empowerment also have to be critically appraised. Depending on the specific context of the problem to be solved, these limitations have been handled to some extent through a clear dichotomy in the blockchain architectures, namely by conceiving both permissioned and permissionless blockchains. Permissionless blockchain requires massive computing power to achieve consensus, while its permissioned counterpart is energy efficient but would require trusted participants. To combine these benefits by gaining trust at a high energy efficiency, a novel mechanism is proposed for automatically learning the trust level of users in a public blockchain network and granting them access to a private blockchain network. In this context, machine learning is a very powerful tool capable of automatically learning the trust level. We have proposed reinforcement learning for bridging the dichotomy of blockchains in terms of striking a trust vs complexity trade-off in an unknown environment. Benefits and limitations of various forms of blockchain techniques are analyzed, followed by their reinforcement-aided evolution. We demonstrate that the proposed reinforcement learning aided blockchain is capable of supporting high-integrity autonomous operation and decision making in wireless networks. The win-win amalgamation of these techniques has been demonstrated for striking a compelling balance between the benefits of permissioned and permissionless blockchain networks through the case-study of the proposed blockchain based unmanned aerial vehicle aided wireless networks.

Multiuser Precoded MIMO Visible Light Communication Systems Enabling Spatial Dimming

In this article, a new indoor multiuser (MU) multiple-input–multiple-output (MIMO) visible light communication (VLC) system is proposed, where the dynamic range of the light emitting diodes (LED-DR) and illumination requirements are taken into consideration. Exploiting the characteristics of LED-DR and the block diagonalization precoding (BDP) technique, we design a weight-adjusted BDP (WA-BDP) scheme that can exploit the LED-DR to offer a high performance in terms of bit error rate (BER). Moreover, utilizing the LED array equipped with a number of LEDs, a spatial dimming (SD) mechanism based on LED selection is employed to fulfill the uniform illumination requirement for typical indoor scenarios. Finally, we combine the WA-BDP technique and SD to form the proposed spatially-dimmed, weight-adjusted block diagonalization precoded (SD-WA-BDP) MU-MIMO-VLC system, where the LED-DR is fully utilized to support simultaneous illumination and communication. In addition, two options of the SD-WA-BDP scheme are developed to manage the performance and complexity tradeoff. Simulation results demonstrate the superiority of the proposed scheme.

Revisiting Efficient Multi-Step Nonlinearity Compensation With Machine Learning: An Experimental Demonstration

Efficient nonlinearity compensation in fiber-optic communication systems is considered a key element to go beyond the "capacity crunch''. One guiding principle for previous work on the design of practical nonlinearity compensation schemes is that fewer steps lead to better systems. In this paper, we challenge this assumption and show how to carefully design multi-step approaches that provide better performance--complexity trade-offs than their few-step counterparts. We consider the recently proposed learned digital backpropagation (LDBP) approach, where the linear steps in the split-step method are re-interpreted as general linear functions, similar to the weight matrices in a deep neural network. Our main contribution lies in an experimental demonstration of this approach for a 25 Gbaud single-channel optical transmission system. It is shown how LDBP can be integrated into a coherent receiver DSP chain and successfully trained in the presence of various hardware impairments. Our results show that LDBP with limited complexity can achieve better performance than standard DBP by using very short, but jointly optimized, finite-impulse response filters in each step. This paper also provides an overview of recently proposed extensions of LDBP and we comment on potentially interesting avenues for future work.

Programming and simulating chemical reaction networks on a surface.

Models of well-mixed chemical reaction networks (CRNs) have provided a solid foundation for the study of programmable molecular systems, but the importance of spatial organization in such systems has increasingly been recognized. In this paper, we explore an alternative chemical computing model introduced by Qian & Winfree in 2014, the surface CRN, which uses molecules attached to a surface such that each molecule only interacts with its immediate neighbours. Expanding on the constructions in that work, we first demonstrate that surface CRNs can emulate asynchronous and synchronous deterministic cellular automata and implement continuously active Boolean logic circuits. We introduce three new techniques for enforcing synchronization within local regions, each with a different trade-off in spatial and chemical complexity. We also demonstrate that surface CRNs can manufacture complex spatial patterns from simple initial conditions and implement interesting swarm robotic behaviours using simple local rules. Throughout all example constructions of surface CRNs, we highlight the trade-off between the ability to precisely place molecules and the ability to precisely control molecular interactions. Finally, we provide a Python simulator for surface CRNs with an easy-to-use web interface, so that readers may follow along with our examples or create their own surface CRN designs.

Receive Beamforming and Resource Allocation for Wireless Powered Non-Orthogonal Multiple Access

This article investigates a wireless powered communication network with non-orthogonal multiple access for collecting data from battery-limited internet-of-things (IoT) devices. To allow various trade-offs in performance and complexity, we design a generalized multi-antenna successive interference cancellation receiver with receive beamforming (BF), power allocation (PA) to devices, and time allocation (TA) optimized toward maximum fairness. We solve the problem in two steps by optimizing the receive BF and PA for a fixed TA and subsequently optimizing the TA. An iterative algorithm and its suboptimal approach are proposed for the BF and PA while a quasiconcave optimization method is adopted for the TA problem by showing its quasiconcave property. Results show that the proposed methods outperform the conventional one with a larger gain for a higher transmit power and greater number of devices.

A Soft-Output STBC Decoder for Aeronautical Telemetry

Alamouti encoding is a well-known space time block encoding technique used to improve the received signal quality in Rayleigh fading channels. In aeronautical telemetry, this encoding technique is applied to shaped offset quadrature phase shift keying tier generation (SOQPSK-TG) modulation in order to handle the two-antenna issue. It is provided for in telemetry-related IRIG standards. In this paper, we propose a unique decoding architecture for Alamouti-encoded SOQPSK-TG signals, taking advantage of pulse amplitude modulation decomposition with soft and hard outputs. We exploit this result to obtain a Viterbi algorithm (VA) for hard decoding and a soft output Viterbi algorithm (SOVA) for soft and hard decoding, with a twofold benefit: operation using one trellis structure, unlike decoders that are based on the 8-waveforms cross-correlated trellis-coded quadrature modulation (XTCQM) approximation, and very attractive bit error rate performance, as well as a complexity trade-off

Evolutionary Prediction of the Trajectory of a Rosette Momentum Jet Group in Flowing Currents

Yan, X. and Mohammadian, A., 2020. Evolutionary prediction of the trajectory of a rosette momentum jet group in flowing currents. Journal of Coastal Research, 36(5), 1059–1067. Coconut Creek (Florida), ISSN 0749-0208.This study proposes a new approach to predicting the trajectory of a rosette momentum jet group in flowing currents, using multigene genetic programming (MGGP), which is an evolutionary-based artificial intelligence (AI) technique. The MGGP algorithm is used to develop explicit mathematical models that predict the dimensionless coordinates of the jet centerline trajectory as functions of the jet-to-ambient velocity ratios, the Reynolds numbers, the dimensionless jet angle, and the dimensionless travel distance. Experimental data are used to train the models, and the optimal models are identified using the Pareto-optimal approach, based on a performance–complexity trade-off. The same data, and some additional unseen data, are used to assess the performances of the developed models. The results show that the MGGP predictions have a good match with both the training and testing experimental datasets. The best MGGP model is also found to be superior to the best single-gene genetic programming (SGGP) model. This study demonstrates the suitability and capability of the MGGP technique in developing models for predicting the trajectory of a rosette momentum jet group in flowing currents, which can be used in many applications in the field of coastal science and engineering, such as the design of coastal outfall systems and assessment of environmental impacts.

Dimensional scaling of high-speed printed organic transistors enabling high-frequency operation

Printed electronics has promised to deliver low-cost, large-area and flexible electronics for mass-market applications for some time; however, so far one limiting factor has been device performance. Over the last decade, great progress has been made in terms of materials, processing and printing resolution for printed transistors. In this article, we review dimensional scaling of printed organic thin-film transistors, which has enabled high-frequency operation. We review different device architectures that require different dimensions to be scaled with accompanying tradeoffs in performance and complexity. Various printing methods have been used to print scaled transistors. Inkjet and gravure printing have seen the greatest improvements. We will focus on gravure printing here as it not only enables high-resolution features but also high-speed printing for low-cost manufacturing. Operating voltage has been scaled down less aggressively due to difficulties with scaling down the thickness of printed gate dielectrics. The performance of organic semiconductor materials has also improved substantially. When processing the semiconductor, the scaling of other device dimensions needs to be considered to optimize performance. Based on these advances, transistor switching frequency has increased dramatically over the last decade with several reports of high-speed printed inverters operating at high kHz to low MHz frequencies, which are promising results for emerging applications of printed electronics.

Multi-objective Cartesian Genetic Programming optimization of morphological filters in navigation systems for Visually Impaired People

Navigation systems for Visually Impaired People (VIP) have improved in the last decade, incorporating many features to ensure navigation safety. Such systems often use grayscale depth images to segment obstacles and paths according to distances. However, this approach has the common problem of unknown distances. While this can be solved with good quality morphological filters, these might be too complex and power demanding. Considering navigation systems for VIP rely on limited energy sources that have to run multiple tasks, fixing unknown distance areas without major impacts on power consumption is a definite concern. Multi-objective optimization algorithms might improve filters’ energy efficiency and output quality, which can be accomplished by means of different quality vs. complexity trade-offs. This study presents NSGA2CGP, a multi-objective optimization method that employs the NSGA-II algorithm on top of Cartesian Genetic Programming to optimize morphological filters for incomplete depth images used by navigation systems for VIP. Its goal is to minimize output errors and structuring element complexity, presenting several feasible alternatives combining different levels of filter quality and complexity—both of which affect power consumption. NSGA2CGP-optimized filters were deployed into an actual embedded platform, so as to experimentally measure power consumption and execution time. We also propose two new fitness functions based on existing approaches from literature. Results showed improvements in visual quality, performance, speed and power consumption, thanks to our proposed error function, proving NSGA2CGP as a solid method for developing and evolving efficient morphological filters.

The Relation between Tax Complexity and Foreign Direct Investments: Evidence Across Countries

This paper analyzes the association between tax complexity and foreign direct investments (FDI) based on the newly developed Tax Complexity Index (TCI) and its components. For a sample of 15,607 new foreign subsidiaries, we find no association between total tax complexity, as proxied by the TCI, and the location probability. When we decompose the TCI into tax code complexity and tax framework complexity, we find opposing associations. Tax code complexity is positively related to the location probability, while tax framework complexity is negatively related to it. These associations are, for example, driven by the complexity of transfer pricing and loss offset regulations in the tax code and the dimensions guidance, audits, as well as filing and payments, in the tax framework. In additional analyses, we find that the associations are sensitive to certain characteristics, such as country-specific and firm-specific characteristics. For example, the positive tax code association diminishes when tax rates are high. Overall, we are the first to provide empirical evidence on potential cost-benefit tradeoffs of tax complexity for FDI and thereby enhance prior literature, which has primarily focused on the costs of tax complexity.

Online estimation for catalyst activity of acetylene hydrogenation reactor

AbstractIn the ethylene industry, the high purity of the ethylene product depends on hydrogenation in acetylene hydrogenation reactor. Because the catalyst deactivation leads to the moving of the operating point, the operation scheme must be adjusted continually according to the catalyst activity. It is necessary to estimate the catalyst activity online. Based on the discrete dynamic model of the acetylene hydrogenation reactor, the extended Kalman filter (EKF) is used to build the soft sensor for catalyst activity. Considering that EKF involves the large computation costs, we propose a method that estimates the parameters of the time‐varying deactivation kinetics model for the tradeoff of accuracy and complexity. The method is effective to reduce computation complexity of estimation, and simultaneously, the accuracy satisfies the process requirement.

Towards a Novel Generalized Chinese Remainder Algorithm for Extended Rabin Cryptosystem

This paper proposes a number of theorems and algorithms for the Chinese Remainder Theorem, which is used to solve a system of linear congruences, and the extended Rabin cryptosystem, which accepts a key composed of an arbitrary finite number of distinct primes. This paper further proposes methods to relax the condition on the primes with trade-offs in the time complexity. The proposed algorithms can be used to provide ciphertext indistinguishability. Finally, this paper conducts extensive experimental analysis on six large data sets. The experimental results show that the proposed algorithms are asymptotically tight to the existing decryption algorithm in the Rabin cryptosystem with the key composed of two distinct primes while maintaining increased generality.

Deep Private-Feature Extraction

We present and evaluate Deep Private-Feature Extractor (DPFE) , a deep model which is trained and evaluated based on information theoretic constraints. Using the selective exchange of information between a user's device and a service provider, DPFE enables the user to prevent certain sensitive information from being shared with a service provider, while allowing them to extract approved information using their model. We introduce and utilize the log-rank privacy, a novel measure to assess the effectiveness of DPFE in removing sensitive information and compare different models based on their accuracy-privacy trade-off. We then implement and evaluate the performance of DPFE on smartphones to understand its complexity, resource demands, and efficiency trade-offs. Our results on benchmark image datasets demonstrate that under moderate resource utilization, DPFE can achieve high accuracy for primary tasks while preserving the privacy of sensitive information.