In expensive multitask optimization problems (ExMTOPs), multiple complex tasks must be optimized simultaneously under limited computational budgets. Existing approaches, often based on surrogate models, aim to approximate objective functions but struggle to generalize across heterogeneous tasks, depend on task-specific sampling, and require frequent retraining. To address these challenges, we propose the Multifactorial Evolutionary Algorithm–Single Step Generative Model (MFEA-SSG), a meta-learning-inspired framework that learns to generate high-quality solutions across tasks. Inspired by meta-learning, we treat each random shuffle of the decision variables as a unique pseudo-task, training the model on a distribution of these tasks to learn a task-agnostic prior about the structure of elite solutions. This process disrupts task-specific dependencies, allowing the model to learn transferable structures from recomposed samples. We then adopt a diffusion-based generative model to learn the distribution of optimal solutions, enabling knowledge transfer across tasks without directly approximating objective functions. To reduce inference cost, we introduce a student model distilled from the diffusion process. Unlike conventional diffusion models that denoise iteratively, the student generates solutions in a single forward pass, significantly reducing inference time. Comprehensive experiments on both general multitask benchmarks and a real-world protein mutation prediction scenario demonstrate that MFEA-SSG achieves high-quality solutions with fast convergence and low computational cost under limited evaluation budgets, outperforming state-of-the-art general and ExMTOPs algorithms.

In evolutionary multitasking, strategies such as crossover operators and skill factor assignment are critical for effective knowledge transfer. Existing improvements to crossover operators primarily focus on low-dimensional variable combinations, such as arithmetic crossover or partially mapped crossover, which are insufficient for modeling complex high-dimensional interactions. Moreover, static or semi-dynamic crossover strategies fail to adapt to the dynamic dependencies among tasks. In addition, current Multifactorial Evolutionary Algorithm frameworks often rely on fixed skill factor assignment strategies, lacking flexibility. To address these limitations, this paper proposes the Multifactorial Evolutionary AlgorithmResidual Learning (MFEA-RL) method based on residual learning. The method employs a Very Deep Super-Resolution (VDSR) model to generate high-dimensional residual representations of individuals, enhancing the modeling of complex relationships within dimensions. A ResNet-based mechanism dynamically assigns skill factors to improve task adaptability, while a random mapping mechanism efficiently performs crossover operations and mitigates the risk of negative transfer. Theoretical analysis and experimental results show that MFEA-RL outperforms state-of-the-art multitasking algorithms. It excels in both convergence and adaptability on standard evolutionary multitasking benchmarks, including CEC2017-MTSO and WCCI2020-MTSO. Additionally, its effectiveness is validated through a real-world application scenario.

Abstract To effectively extract and classify the information from reports or documents and protect the privacy of the extracted results, we propose a privacy classification named Word Embedding Combination Privacy-preserving Support Vector Machine (WECPPSVM) model to classify the text. In addition, this paper also proposes the Privacy-preserving Distribution and Independent Frequent Subsequence Extraction Algorithm (PPDIFSEA), which calculates the degree of independence of the training data input to the classification model by training the Deep Belief Network(DBN) in PPDIFSEA, then obtains the Privacy Boundary(PB). PB is an indispensable condition for both data sampling and privacy noise generation. And this model can protect privacy by injecting the privacy noise into the classification result, this method can interfere with the background knowledge-based privacy attack. Our quantitative analysis shows that the WECPPSVM proposed in this paper can approach mainstream text classification algorithms in terms of text classification accuracy while preserving privacy without increasing computational complexity. In addition, the fusion study and privacy threat evaluation also verify that the proposed PPDIFSEA method combined with WECPPSVM achieves an acceptable level of classification accuracy and privacy protection.

Quantification research has sought to accurately estimate class distributions under dataset shift. While existing methods perform well under assumed conditions of shift, it is not always clear whether such assumptions will hold in a given application. This work extends the analysis and experimental evaluation of our Gain-Some-Lose-Some (GSLS) model for quantification under general dataset shift and incorporates it into a method for dynamically selecting the most appropriate quantification method. Selection by a Kolmogorov-Smirnov test for any shift followed by a newly proposed “Adjusted Kolmogorov-Smirnov” test for non-prior shift is found to best balance quantification and runtime performance. We also present a framework for constraining quantification prediction intervals to user-specified limits by requesting a smaller set of instance class labels from the user than required with confidence-based rejection.

In this paper, we present a novel Liquid State Machine (LSM) based approach for modelling of multimodal longitudinal data: the Mosaic LSM. Our model harnesses the strengths of multiple LSMs, each designed to capture the temporal patterns of a specific data modality. This temporal information is then added to the raw data to create a composite representation that encompasses both the multimodal and the longitudinal aspects of the data. We demonstrate the performance of our approach on a real-world dataset that contains clinical, cognitive, and genetic modalities with the aim of predicting the Ultra-High Risk (UHR) status in individuals, six months in advance. Our results show that the Mosaic LSM outperforms traditional machine learning models, achieving an outstanding Matthew's Correlation Coefficient of 0.84 and prediction accuracy of 92.4%. Overall, our work highlights the potential of Mosaic LSM as a powerful tool for disease prognosis, and its ability to leverage both the multimodality and temporality of the data to improve performance.

Selecting informative features, such as accurate biomarkers for disease diagnosis, prognosis and response to treatment, is an essential task in the field of bioinformatics. Medical data often contain thousands of features and identifying potential biomarkers is challenging due to small number of samples in the data, method dependence and non-reproducibility. This paper proposes a novel ensemble feature selection method, named Filter and Wrapper Stacking Ensemble (FWSE), to identify reproducible biomarkers from high-dimensional omics data. In FWSE, filter feature selection methods are run on numerous subsets of the data to eliminate irrelevant features, and then wrapper feature selection methods are applied to rank the top features. The method was validated on four high-dimensional medical datasets related to mental illnesses and cancer. The results indicate that the features selected by FWSE are stable and statistically more significant than the ones obtained by existing methods while also demonstrating biological relevance. Furthermore, FWSE is a generic method, applicable to various high-dimensional datasets in the fields of machine intelligence and bioinformatics.

In recent years, various approaches for multiresident human activity recognition (HAR) in a smart indoor environment have been developed and improved along with the rapid development of sensors and AI technologies. Research in data stream-based online learning (OL) for multiresident HAR is relatively new and a majority of the existing works have been developed based on training batches of data that cannot recognize real-time activities. To address the challenges of OL for multiresident HAR, we propose a novel OL architecture based on a locally weighted ensemble detection-based adaptive random forest (LED-ARF) classifier. We conduct a comprehensive performance comparison of eight famous OL classification techniques and our LED-ARF method. The comparison is evaluated based on the two benchmarking CASAS and ARAS data sets. Our experimental results show that LED-ARF achieves the best performance with the highest robustness for online multiresident HAR.

Several techniques for human activity recognition (HAR) in a smart indoor environment have been developed and improved along with the rapid advancement of sensor technologies. However, recognizing multiple people’s activities is still challenging due to the complexity of their activities, such as parallel and collaborative activities. To address these challenges, we propose a transformer with a bidirectional gated recurrent unit (GRU) deep learning (DL) method, called TRANS-BiGRU, to efficiently learn and recognize different types of activities performed by multiple residents. We compare the proposed model with the state-of-the-art models and various DL models, such as Ensemble2LSTM (Ens2-LSTM), bidirectional GRUs (Bi-GRU), and traditional machine learning (ML) models, such as support vector machine (SVM). Our experimental results based on the center for advanced studies in adaptive system and ARAS public data sets show that our model significantly outperforms the existing models for complex activity recognition of multiple residents.

Abstract Attention mechanisms have been incorporated into many neural network-based natural language processing (NLP) models. They enhance the ability of these models to learn and reason with long input texts. A critical part of such mechanisms is the computation of attention similarity scores between two elements of the texts using a similarity score function. Given that these models have different architectures, it is difficult to comparatively evaluate the effectiveness of different similarity score functions. In this paper, we proposed a baseline model that captures the common components of recurrent neural network-based Question Answering (QA) systems found in the literature. By isolating the attention function, this baseline model allows us to study the effects of different similarity score functions on the performance of such systems. Experimental results show that a trilinear function produced the best results among the commonly used functions. Based on these insights, a new T-trilinear similarity function is proposed which achieved the higher predictive EM and F1 scores than these existing functions. A heatmap visualization of the attention score matrix explains why this T-trilinear function is effective.

Effective supervised training of modern machine learning models often requires large labelled training datasets, which could be prohibitively costly to acquire for many practical applications. Research addressing this problem has sought ways to leverage weak supervision sources, such as the user-defined heuristic labelling functions used in the data programming paradigm, which are cheaper and easier to acquire. Automatic generation of these functions can make data programming even more efficient and effective. However, existing approaches rely on initial supervision in the form of small labelled datasets or interactive user feedback. In this paper, we propose Witan, an algorithm for generating labelling functions without any initial supervision. This flexibility affords many interaction modes, including unsupervised dataset exploration before the user even defines a set of classes. Experiments in binary and multi-class classification demonstrate the efficiency and classification accuracy of Witan compared to alternative labelling approaches.

In order to perform competitive privacy-guaranteed object detection, we propose an end-to-end model called Privacy-preserving Deep Transformation Self-attention (PPDTSA). This model ensures the privacy of the inference results. It has a low-complexity hierarchical structure with a relatively small number of hyper-parameters. Consistency of prediction is achieved through the encoding and decoding blocks of the self-attention mechanism which enables points of interest to be located. Focus loss is estimated based on foreground-background imbalance. The remaining dense blocks enable image details to be retained and the Region Of Interest to be expanded. At the same time, the objects detected in the image are protected through the privacy noise volume which is specified by the user. Experimental results demonstrate that PPDTSA achieves superior performance on the MOT20 dataset compared with three other state-of-the-art object detection models.

When applying supervised learning to estimate class distributions of unlabelled samples (so-called quantification), dataset shift is an expected yet challenging problem. Existing quantification methods make strong assumptions on the nature of dataset shift that often will not hold in practice. We propose a novel Gain-Some-Lose-Some (GSLS) model that accounts for more general conditions of dataset shift. We present a method for fitting the GSLS model without any labelled instances from the target sample, and experimentally demonstrate that GSLS can produce reliable quantification prediction intervals under broader conditions of shift than existing quantification methods.

Multiresident activity recognition (AR), which has become a popular research field in smart environments, aims to recognize the activities of multiple residents based on data collected from various types of sensors, and sensor events segmentation is an important technique for enhancing the performance of AR. While quite some segmentation methods have been proposed for the single person setting, few studies have been done for the multiresident setting. In this article, we first evaluate the baseline and the state-of-the-art segmentation methods using the popular multiresident data set CASAS, to confirm that the performance of multiresident AR can be improved by applying segmentation techniques; we then propose a novel Hybrid fuzzy c-means (FCM) change point detection (CPD)-based segmentation method that can further enhance the performance of multiresident AR. We combine a FCM method with a CPD-based method for sensor event segmentation. The FCM method is used to classify the sensor events in terms of sensor locations, and then the CPD technique is used to probe the transition actions to determine the segmentation sequence. Our experimental results show that the proposed method significantly improves the performance of multiresident AR in comparison with the baseline and state-of-the-art classification methods.

In this paper, we present a Convolutional AutoEncoder (CAE) for single image dehazing. Our CAE makes use of Densely Connection Networks as its encoder and decoder. It is trained with the corresponding hazy and clean images at the input and output, enabling it to remove the haze without having to rely on an atmospheric scattering model. The CAE is trained and tested with the RESIDE dataset. Experiment results show that this CAE outperforms eight state-of-art methods. The trained CAE is also applied to some real-life hazy images, and decent dehazing results are obtained. Moreover, our method is computationally efficient enough to run on computers without GPU units.

A human-swarm interactive system has been shown to benefit from neglect benevolence. This means if the human interacts too often with a swarm of robots, it will degrade its performance in achieving the operational goal. Past works have focused on human-swarm interaction (HSI) for reaching rendezvous points. In this paper, we aim to establish neglect benevolence for realignment in a leader-follower Cucker-Smale system. The human operator issues heading commands to the leader and the follower robots tries to realign with the leader through Cucker-Smale model dynamics. Simulation results show that there is a minimum time interval that the operator should wait before issuing a new heading command in order for the swarm to remain in a flocking state.

It has been observed empirically that a flock of birds is able to turn and change direction of flight rapidly as a single unit. The “chorus-line hypothesis” was proposed in 1984 by Potts to explain this phenomenon. It puts forward the idea that those birds that are further down the line from the initiator can predict the movement from those nearer the initiator, thereby reducing their response time. Inspired by this biological phenomenon, this paper proposes an implementation of the chorus-line effect in a Cucker-Smale multiple agent system. We study the time it takes for the rest of the agents to follow an abrupt change in direction executed by one of the flock members. Through mathematical analysis and computer simulation, we found that the time required for this kind of manoeuvre tracking can indeed be reduced compared with the standard Cucker-Smale model. It is more effective than using finite-time control that has been studied earlier for smaller flocks. Furthermore, using finite-time control on the chorus-line incorporated model produces the shortest realignment time for all flock sizes considered.

Convolved Gaussian process (CGP) is a type Gaussian process modelling technique applicable for multiple-input multiple-output systems. It employs convolution processes to construct a covariance function that models the correlation between outputs. Modelling using CGP involves learning the hyperparameters of the latent function and the smoothing kernel. Conventionally, learning involves the maximization of the log likelihood function of the training samples using conjugate gradient (CG) or particle swarm optimization (PSO) methods. We propose to use PSO to minimize the model error. In this way, a clearer direct indication of the quality of the current solution during the optimization process can be obtained. Simulation results on a dynamical system show that our method is able to learn appropriate CGP models and achieve better predictive performance compared with CG when the searching space is not well defined.

This brief presents a new low-complexity reconfigurable fast filter bank (RFFB) for wireless communication applications such as spectrum sensing and channelization. In RFFB, the bandwidth and center frequency of sub-bands can be varied with high frequency resolution without hardware reimplementation. This is achieved with an improved modified frequency transformation-based variable digital filter (MFT-VDF) at the first stage of the proposed multistage implementation. Existing second-order frequency transformation-based low-pass VDFs have limited cutoff frequency range which is approximately 12.5% of the sampling frequency. The proposed low-pass MFT-VDF offers unabridged control over the cutoff frequency on a wide frequency range thereby, improving the cutoff frequency range of existing VDFs. The design example shows that the RFFB is easy to design and offers substantial savings in gate counts over other filter banks.

In this brief, an efficient implementation of reconfigurable warped digital filter with variable low-pass, high-pass, bandpass, and bandstop responses is presented. The warped filters, obtained by replacing each unit delay of a digital filter with an all-pass filter, are widely used for various audio processing applications. However, warped filters require first-order all-pass transformation to obtain variable low-pass or high-pass responses, and second-order all-pass transformation to obtain variable bandpass or bandstop responses. To overcome this drawback, the proposed method combines the warped filters with the coefficient decimation technique. The proposed architecture provides variable low-pass or high-pass responses with fine control over cut-off frequency and variable bandwidth bandpass or bandstop responses at an arbitrary center frequency without updating the filter coefficients or filter structure. The design example shows that the proposed variable digital filter is simple to design and offers substantial savings in gate counts and power consumption over other approaches.

This paper presents the design of a variable linear phase finite impulse response filter based on second order frequency transformations and coefficient decimation. The design of variable digital filters (VDFs) using first and second order frequency transformations have been proposed in literature. The VDF using second order transformation has better cut-off slope characteristics compared to the VDF using first order transformation. However, the former has the drawback of limited range (approximately 25% of the half of the sampling frequency) over which the cut-off frequency, fc, can be varied. It also fails to provide variable lowpass, highpass, bandpass or bandstop responses from a fixed-coefficient lowpass filter using the same architecture. The architecture proposed here overcomes the above mentioned disadvantages using coefficient decimation technique. The design example shows that the range over which fccan be varied is 2.65 times wider in the proposed VDF than the VDF in [7] and for a given frequency range, the proposed VDF offers a total gate count saving of 33% and 41% over the VDF in [11] and [7] respectively. Also, the proposed architecture provides variable lowpass, highpass, bandpass or bandstop responses from a fixed coefficient lowpass filter.

This paper presents a new method for the design of finite impulse response (FIR) filter that provides variable frequency responses. The proposed idea is to replace each unit delay operator in a fixed-coefficient FIR filter with the 2ndorder FIR fractional delay (FD) structure and the cutoff frequency, fcof the filter is changed by changing the FD value. The change in FD results in change in amplitude and length of an impulse response. This in turn changes fcand transition bandwidth (TBW) of an FIR filter. The mathematical relation between cut-off frequency, TBW and FD value D is derived. The design example shows that the proposed method provides very fine control over fc.

Therapeutically, the closed-loop blood glucose-insulin regulation paradigm via a controllable insulin pump offers a potential solution to the management of diabetes. However, the development of such a closed-loop regulatory system to date has been hampered by two main issues: 1) the limited knowledge on the complex human physiological process of glucose-insulin metabolism that prevents a precise modeling of the biological blood glucose control loop; and 2) the vast metabolic biodiversity of the diabetic population due to varying exogneous and endogenous disturbances such as food intake, exercise, stress, and hormonal factors, etc. In addition, current attempts of closed-loop glucose regulatory techniques generally require some form of prior meal announcement and this constitutes a severe limitation to the applicability of such systems. In this paper, we present a novel intelligent insulin schedule based on the pseudo self-evolving cerebellar model articulation controller (PSECMAC) associative learning memory model that emulates the healthy human insulin response to food ingestion. The proposed PSECMAC intelligent insulin schedule requires no prior meal announcement and delivers the necessary insulin dosage based only on the observed blood glucose fluctuations. Using a simulated healthy subject, the proposed PSECMAC insulin schedule is demonstrated to be able to accurately capture the complex human glucose-insulin dynamics and robustly addresses the intraperson metabolic variability. Subsequently, the PSECMAC intelligent insulin schedule is employed on a group of type-1 diabetic patients to regulate their impaired blood glucose levels. Preliminary simulation results are highly encouraging. The work reported in this paper represents a major paradigm shift in the management of diabetes where patient compliance is poor and the need for prior meal announcement under current treatment regimes poses a significant challenge to an active lifestyle.

A new approach to implement computationally efficient reconfigurable filter banks for multi-standard wireless receivers is presented in this paper. Based on the concepts of tree-structured quadrature mirror filter bank (TQMFB) and coefficient decimation approach, a reconfigurable and efficient tree-structured non-uniform filter bank (TNFB) is proposed in this paper. The proposed filter bank is designed to extract channels of non-uniform bandwidths with reduced complexity when compared to TQMFB. Each stage of the proposed filter bank consists of a modal filter and a complementary delay to obtain the low-pass and high-pass channels respectively. Design examples show that the proposed TNFB offers an average multiplication rate reduction of 69% over the TQMFB.

The cerebellum constitutes a vital part of the human brain system that possesses the capability to model highly nonlinear physical dynamics. The cerebellar model articulation controller (CMAC) associative memory network is a computational model inspired by the neurophysiological properties of the cerebellum, and it has been widely used for control, optimization, and various pattern recognition tasks. However, the CMAC network's highly regularized computing structure often leads to the following: 1) a suboptimal modeling accuracy, 2) poor memory utilization, and 3) the generalization-accuracy dilemma. Previous attempts to address these shortcomings have limited success and the proposed solutions often introduce a high operational complexity to the CMAC network. This paper presents a novel neurophysiologically inspired associative memory architecture named pseudo-self-evolving CMAC (PSECMAC) that nonuniformly allocates its computing cells to overcome the architectural deficiencies encountered by the CMAC network. The nonuniform memory allocation scheme employed by the proposed PSECMAC network is inspired by the cerebellar experience-driven synaptic plasticity phenomenon observed in the cerebellum, where significantly higher densities of synaptic connections are located in the frequently accessed regions. In the PSECMAC network, this biological synaptic plasticity phenomenon is emulated by employing a data-driven adaptive memory quantization scheme that defines its computing structure. A neighborhood-based activation process is subsequently implemented to facilitate the learning and computation of the PSECMAC structure. The training stability of the PSECMAC network is theoretically assured by the proof of its learning convergence, which will be presented in this paper. The performance of the proposed network is subsequently benchmarked against the CMAC network and several representative CMAC variants on three real-life applications, namely, pricing of currency futures option, banking failure classification, and modeling of the glucose-insulin dynamics of the human glucose metabolic process. The experimental results have strongly demonstrated the effectiveness of the PSECMAC network in addressing the architectural deficiencies of the CMAC network by achieving significant improvements in the memory utilization, output accuracy as well as the generalization capability of the network.

This paper is focus on to design a sampling system with minimum sampling rate for signals in the shift invariant Hilbert space. To achieve this goal, we propose a method to calculate the rate of innovation (RI) of signals in the Hilbert space. The RI is then used to identify a suitable sampling kernel as well as the sampling rate for a specific signal. We show that the RI of the kernel should be greater or equal to the RI of the signal for the signal to be perfectly reconstructible. The minimum sampling rate depends on the RI of the signal. Examples are included to demonstrate how our method is applied to calculate the RI. Some known sampling theories can also be fitted into the framework of sampling system with minimum sampling rate.

The complexity of finite impulse response (FIR) filters is dominated by the number of adders (subtractors) used to implement the coefficient multipliers. A greedy common subexpression elimination (CSE) algorithm with a look-ahead method based on the canonic signed digit (CSD) representation of filter coefficients for implementing low complexity FIR filters is proposed in this paper. Our look-ahead algorithm chooses the maximum number of frequently occurring common subexpressions and hence reduces the number of adders required to implement the filter. This adder reduction is achieved without any increase in critical path length. Design examples of FIR filters show that the proposed method offers an average adder reduction of about 20% over the best known CSE method.

The focus of this paper is on the minimization of the variation in power consumed by a VLIW processor during the execution of a target program through instruction scheduling. The problem is formulated as a mixed-integer program (MIP) and a problem-specific branch-and-bound algorithm has been developed to solve it more efficiently than generic MIP solvers. Simulation results based on the TMS320C6711 VLIW digital signal processor using benchmarks from Mediabench and Trimaran showed that over 40% average reduction in power variation can be achieved without sacrificing execution speed of these benchmarks. Computational requirements and convergence rates of our algorithm are also analyzed.

The cerebellar model articulation controller (CMAC) neural network (NN) is a well-established computational model of the human cerebellum. Nevertheless, there are two major drawbacks associated with the uniform quantization scheme of the CMAC network. They are the following: (1) a constant output resolution associated with the entire input space and (2) the generalization-accuracy dilemma. Moreover, the size of the CMAC network is an exponential function of the number of inputs. Depending on the characteristics of the training data, only a small percentage of the entire set of CMAC memory cells is utilized. Therefore, the efficient utilization of the CMAC memory is a crucial issue. One approach is to quantize the input space nonuniformly. For existing nonuniformly quantized CMAC systems, there is a tradeoff between memory efficiency and computational complexity. Inspired by the underlying organizational mechanism of the human brain, this paper presents a novel CMAC architecture named hierarchically clustered adaptive quantization CMAC (HCAQ-CMAC). HCAQ-CMAC employs hierarchical clustering for the nonuniform quantization of the input space to identify significant input segments and subsequently allocating more memory cells to these regions. The stability of the HCAQ-CMAC network is theoretically guaranteed by the proof of its learning convergence. The performance of the proposed network is subsequently benchmarked against the original CMAC network, as well as two other existing CMAC variants on two real-life applications, namely, automated control of car maneuver and modeling of the human blood glucose dynamics. The experimental results have demonstrated that the HCAQ-CMAC network offers an efficient memory allocation scheme and improves the generalization and accuracy of the network output to achieve better or comparable performances with smaller memory usages. Index Terms-Cerebellar model articulation controller (CMAC), hierarchical clustering, hierarchically clustered adaptive quantization CMAC (HCAQ-CMAC), learning convergence, nonuniform quantization.

The resampling of discrete-time signals where the underlying analog signal is non-bandlimited is considered in this paper. We extend the generalized sampling theory developed based on the principle of consistency to resampling. Realizing the resampling system has both discrete input and output, the performance of the resampling filter is considered in l2instead of the traditionally used L2. We show that the performance of the resampling system depends on the resampling rate instead of the actual interpolating kernels. The theory can be applied to image processing applications like zooming to provide better response to high frequency components. Since the resampling process is discrete in nature, our filter designed to optimize resampling in l2is shown to outperform other techniques designed in L2

The most computationally intensive part of the wideband receiver of a software defined radio (SDR) is the channelizer since it operates at the highest sampling rate. Higher order FIR channel filters are needed in the channelizer to meet the stringent adjacent channel attenuation specifications of wireless communications standards. In this paper, we present a coefficient-partitioning algorithm for realizing low power and high-speed channel filters. Design examples of the channel filters employed in the digital advanced mobile phone system (D-AMPS) and personal digital cellular (PDC) receivers show that the average reductions of memory and power consumption achieved using our method over existing method are 25% and 50% respectively

The most computationally intensive part of the wideband receiver of a software defined radio (SDR) is the intermediate frequency (IF) processing block. Digital filtering is the main task in IF processing. The computational complexity of finite impulse response (FIR) filters used in the IF processing block is dominated by the number of adders (subtracters) employed in the multipliers. This paper presents a method to implement FIR filters for SDR receivers using minimum number of adders. We use an arithmetic scheme, known as pseudo floating-point (PFP) representation to encode the filter coefficients. By employing a span reduction technique, we show that the filter coefficients can be coded using considerably fewer bits than conventional 24-bit and 16-bit fixed-point filters. Simulation results show that the magnitude responses of the filters coded in PFP meet the attenuation requirements of wireless communication standard specifications. The proposed method offers average reductions of 40% in the number of adders and 80% in the number of full adders needed for the coefficient multipliers over conventional FIR filter implementation methods

An instruction word in VLIW (very long instruction word) processors consists of a variable number of individual instructions. Therefore the power consumption variation over time significantly depends on the parallel instruction schedule generated by the compiler. Sharp power variations across time cause power supply noises, degrade chip reliability and accelerate battery exhaustion. This paper proposes a branch and bound algorithm for instruction scheduling of VLIW architectures that effectively minimizing power variation without degrading the speed. Our experimental results demonstrate the efficiency of our algorithm compared with previously presented approaches. Finally, a new rough sets based approach to the instruction-level VLIW power model for this instruction scheduling optimization problem is discussed.

Current techniques for power-aware VLIW instruction scheduling assume that the power consumption parameters are precisely known. In reality, there is always some degree of imprecision. We propose to apply rough set theory to handle the imprecision involved. Power consumption parameters are modeled as rough variables and the power-balanced instruction scheduling problem is formulated as a rough program. The effectiveness and advantages of our approach are illustrated through examples.

The most computationally intensive part of a wideband receiver is the channelizer. The computational complexity of linear phase finite impulse response (LPFIR) filters employed in the channelizer is dominated by the number of adders used in the implementation of the multipliers. In this paper, two methods are proposed to efficiently implement the channel filters in a wideband receiver based on common subexpression elimination (CSE). We exploit the fact that a significant amount of redundant multiplications exist in the filter-bank channelizer as it extracts multiple narrowband channels from the wideband signal. By forming three and four nonzero-bit super-subexpressions utilizing redundant identical shifts that exist between a two- nonzero-bit common subexpression (CS) and a third nonzero bit, or between two nonzero-bit CS, the number of adders to implement the channel filters can be reduced considerably. Furthermore, the complexity of the adders is analyzed and design examples of the channel filters employed in the digital advanced mobile phone system (D-AMPS) and the personal digital cellular (PDC) channelizers show that the proposed methods offer considerable reduction in the number of full adders when compared to conventional CSE methods.

Algorithms that minimize the complexity of multiplication in digital filters focus on reducing the number of adders needed to implement the coefficient multipliers. Previous works have not analyzed the complexity of each adder, which is significant in low-complexity implementation. A multiplication algorithm for low-complexity implementation of digital filters with a minimum number of full adders (NFAs) and improved speed is proposed here. The authors exploit the fact that when multiplication is implemented using shifts and adds, the adder width can be minimized by limiting the shifts of the operands to shorter lengths. The coefficient-partitioning (CP) algorithm proposed here minimizes the shifts of the operands of the adders by partitioning each coefficient into two subcomponents. The authors show that by combining three methods, the CP algorithm, an efficient coefficient coding scheme known as pseudo floating-point (PFP) representation, and the well-known common subexpression elimination (CSE), the NFAs required in each adder of the multiplier can be reduced considerably. Design examples show that the method offers an average FA reduction of 30% for finite-impulse response (FIR) filters and 20% for infinite-impulse response (IIR) filters over CSE methods.

The objective of audio source separation is to separate sound mixtures into individual streams based on the sources. It has many potential applications, one of which is in a system for perceptually-based search and retrieval of audio data from multimedia databases. The task of audio source separation is very difficult if all audio data are mixed into a single track. In this paper, we restrict the single track recordings to instrumental music. Hence, we attempt to construct separate streams of data each consisting of the sound of a single instrument. A model-based approach is used. The architecture of the system is based on a cerebellar-based (CMAC) fuzzy neural network. The subjective test results of our experiments on the separation of 2-source audio mixtures show that our approach is promising.

The computational complexity of linear phase finite impulse response (LPFIR) filters used in the channelizer of a wideband receiver is dominated by the number of adders (subtracters) employed in the multipliers. Common subexpression elimination (CSE) is a well-known technique for minimizing the number of adders in LPFIR filters. An improved CSE method is proposed in this paper, which is used to implement the channel filters of a filter bank channelizer (FBC). In the FBC, each modulated bandpass filter extract one channel from the input wideband signal. The reduction in number of adders is obtained by eliminating redundant multiplications of common subexpressions that exist among the channel filters of the FBC with the input signal. Design example of the channel filters employed in the digital advanced mobile phone system (D-AMPS) show that the proposed method offers considerable reduction in the number of full adders when compared with conventional CSE methods.

The flexibility of a software-defined radio (SRR) depends on its capability to operate in multi-standard wireless communication environments. The most computationally intensive part of wideband receivers is the channelizer, which extracts multiple narrowband signals from adjacent frequency hands. In an SDR receiver, the compatibility of the channelizer with different communication standards is guaranteed by its reconfigurability. This paper presents an efficient channelizer that has a reconfigurable architecture based on quadrature mirror filter bank (QMF) trees. We show that the channelizer can he efficiently implemented using common subexpression based filter structures. An example of dual-mode global system for mobile communication (GSM)/personal digital cellular (PDC) channelizer is discussed to illustrate the proposed design methodology.

The paper describes an intelligent computer aided architectural design system (ICAAD) called ICADS. ICADS encapsulates different types of design knowledge into independent "critic" modules. Each "critic" module possesses expertise in evaluating an architect's work in different areas of architectural design and can offer expert advice when needed. This research focuses on the representation of spatial information encoded in architectural floor plans and the representation of expert design knowledge. Described in the paper is our research in designing and developing two particular "critic" modules. The first module, FPDX, checks a residential apartment floor plan, verifies that the plan meets a set of government regulations, and offers suggestions for floor plan changes if regulations are not met. The second module, IDX, analyzes room and furniture layout according to a set of interior design guidelines and offers ideas on how furniture should be moved if the placement does not follow good design principles.

In this paper, a bidirectional associative memory (BAM) model with second-order connections, namely second-order bidirectional associative memory (SOBAM), is first reviewed. The stability and statistical properties of the SOBAM are then examined. We use an example to illustrate that the stability of the SOBAM is not guaranteed. For this result, we cannot use the conventional energy approach to estimate its memory capacity. Thus, we develop the statistical dynamics of the SOBAM. Given that a small number of errors appear in the initial input, the dynamics shows how the number of errors varies during recall. We use the dynamics to estimate the memory capacity, the attraction basin, and the number of errors in the retrieved items. Extension of the results to higher-order bidirectional associative memories is also discussed.

The stability, capacity and statistical dynamics of second-order bidirectional associative memory (BAM) are presented here. We first use an example to illustrate that the state of second-order BAR I may converge to limited cycles. When error in the retrieved pairs is not allowed, a lower bound of memory capacity is derived. That is O(min(n/sup 2//(log n),p/sup 2//(log p))) where n and p are the dimensions of the library pairs. Since the state of second-order BAM may converge to limited cycles, the conventional method cannot be used to estimate its memory capacity when small errors in the retrieval pairs are allowed. Hence, the statistical dynamics of second-order BAM is introduced: starting with an initial state close to the library pairs, how the confidence interval of the number of errors changes during recalling. From the dynamics, the attraction basin, memory capacity, and final error in the retrieval pairs can be estimated. Also, some numerical results are given. Finally, an extension of the results to higher-order BAM is discussed.>

The authors present a report on the content and status of a major speech database collection effort. The goal is to collect a useful set of speech material from a very large number of speakers. These speakers are drawn from a wide cross-section of the local community with a variety of ethnic and educational backgrounds. Speech materials include isolated digits and numbers, vowels and voiced phonemes, connected digits, and phonetically balanced sentences. Speech signals are encoded into 16-bit pulse code modulation (PCM) format and stored on Betamax format video tapes. In the seven months since this project started, speech from 100 speakers has been collected. A statistical breakdown of the backgrounds of the speakers is presented.>