Noisy recurrent neural networks: the continuous-time case

The classical stochastic analog of the deterministic linear system in engineering is the linear system driven by white noise. As the promise of artificial neural networks in modeling nonlinear systems continues to grow, the need for a stochastic analog with quantitative foundations for analysis and synthesis will increase. This paper presents recent work in this direction, examining recurrent neural nets (RNN driven by white noise. We examine the effect of noise on the typical continuous-time RNN model. First, we perform qualitative analysis establishing uniform boundedness of moments of the neuron states over time. To enable practical application, however, it is necessary to relate these properties to useful measures that can be estimated. We thus subsequently derive bias and variance measures for the noisy RNN with respect to the corresponding deterministic RNN. This has significant practical implications, since net design is nonminimal in the sense that several nets can solve the same problem. The results allow the user to evaluate given RNN for noise performance. The designer can use these results to constrain the design space so that the design satisfies performance specifications whenever possible. An example is provided using the measures derived in this paper to predetermine the best among several RNN designs for a given problem.     

The Cognitive Benefit of Dynamic Representations on Procedural Skill Acquisition: A Computational Modeling Approach

Cognitive computational modeling is a viable methodology for further investigation of the hitherto inconclusive findings on the cognitive benefits of dynamic versus static visualization components of instructions. This is more so as contemporary cognitive architectures such as the Adaptive Control of Thought–Rational (ACT–R) 6.0 are increasingly applied to traditional cognitive psychology research problems. The application of this methodology is, however, restricted by the limited capability of existing architectures for implementing detailed atomic motor actions such as those involved in complex skill acquisition and performance. This article presents a 2-component computational modeling methodology for investigating the cognitive processes involved in the acquisition and performance of skilled motor tasks. The approach specifies a novel combination of a sequence-of-point technique with a movement control mechanism to implement variously acquired cognitive mental task representations and their intertwined role in postlearning performance as evident in the atomic control of motor actions. This paradigm is validated for 2 experiments using incrementally developed cognitive models developed in ACT–R 6.0. The model's quantitative outputs correlate significantly with equivalent empirical human data. This has implications for multimedia instructional design, especially where rapid, transferrable skill acquisition is desired on initial exposure.     

Direct adaptive neural control of antilock braking systems incorporated with passive suspension dynamics

A direct adaptive neural network-based feedback linearization (NNFBL) slip control scheme for an antilock braking system (ABS) is presented. The NNFBL slip controller is developed to minimise the vehicle braking distance and to simultaneously improve its overall ride comfort and road handling. The comprehensive vehicle model incorporates the passive suspension dynamics, the dynamics of the electro-mechanical based braking system and air drag and wheel bearing friction. A feedforward, multilayer perceptron (MLP) neural network (NN) model that is well suited for control by discrete input-output linearization (NNIOL) is selected to represent the ABS with passive suspension. The NN model was trained using Levenberg-Marquardt optimization algorithm. The controlled signal was further boosted using a genetic algorithm generated gain. The effectiveness of the proposed controller is demonstrated by simulation results, in the presence of deterministic road disturbance input to the suspension and varying road conditions. The results are superior with respect to braking distance minimization and also to reference slip tracking, especially on the dry asphalt road.     

Recurrent neural network training with feedforward complexity

This paper presents a training method that is of no more than feedforward complexity for fully recurrent networks. The method is not approximate, but rather depends on an exact transformation that reveals an embedded feedforward structure in every recurrent network. It turns out that given any unambiguous training data set, such as samples of the state variables and their derivatives, we need only to train this embedded feedforward structure. The necessary recurrent network parameters are then obtained by an inverse transformation that consists only of linear operators. As an example of modeling a representative nonlinear dynamical system, the method is applied to learn Bessel's differential equation, thereby generating Bessel functions within, as well as outside the training set.     

Kinetics and neuro-fuzzy soft computing modelling of river turbid water coag-flocculation using mango (Mangifera indica) kernel coagulant

This study investigates kinetics and Adaptive Neuro-Fuzzy Modeling (ANFM) of river turbid water coagulation-flocculation (CF) process using mango kernel coagulant (MKC). CF experiments were performed using jar test apparatus and the process kinetic-transport parameters (coagulation rate constant, half-life time, and particle diffusivity) were determined using kinetic-transport models. Grid-partitioning neuro-fuzzy programming codes were written and implemented in Matlab 9.2 software environment for the development of neuro-fuzzy architecture. The ANFM input data include initial water pH, initial water turbidity, biocoagulant dosage, CF time, and turbidity removal percentage (TRP) as output data. Generalized bell membership function was optimally selected for fuzzification of input variables and a hybrid algorithm was considered for the learning method of input-output data with constant output membership type. The minimum turbidity (0.51 NTU) of treated water was achieved at pH 12 and coagulant dosage of 2.5?mg/L with coagulation rate constant, half-life (t_1/2) and particle diffusivity 0.0194?s^?1, 10.01?min, and 7.267?×?10^?14 m²/s, respectively. The correlation coefficient (R²) between the experimental and neuro-fuzzy predicted values was 0.9924 and the ratio (K) of training error to testing error was 0.68. Thus, this study shows that ANFM can be used as a reliable tool for modeling river water CF and kinetic-transport parameter results are useful in process design, optimization, and control.     

From biomimicry to bioelectronics: Smart materials for cardiac tissue engineering

Effective strategies in cardiac tissue engineering require matrices that recapitulate the mechanical, topographic and electrical cues present in the native extracellular matrix. In this review, we discuss recent efforts in materials science and nanotechnology to achieve functional three-dimensional (3D) scaffolds that modulate and monitor cardiac tissue function. We consider key design considerations, including choice of biopolymer matrix, cell sources, and delivery methods for eventual therapies. We then discuss how solid-state nanomaterials may be integrated within these systems to provide unique electrical and nanotopographic cues that improve electromechanical synchrony. We describe how these approaches may be extended to complex, spatially heterogeneous constructs using 3D bioprinting techniques. Finally, we describe how scaffold materials may be augmented with bioelectronic components to achieve hybrid myocardium that monitors or controls electrophysiological activity. Collectively, these approaches provide a framework for achieving cardiac tissues with tunable, rationally-designed functionalities. We discuss future prospects and remaining challenges for clinical translation.

     

Breakthrough in privacy concerns and lawful access conflicts

The widespread use and misuse of communication systems, especially in the era of speedy transmission of audio/visual information facilitated by the World Wide Web, creates a need for the regulation of information flow. This is in order to ensure a high level of consistency in information security and integrity. On the other hand, ensuring that users have access to security techniques that meet their needs, so that they can trust in the security of information and communications systems while maintaining the confidentiality and integrity of data on such systems, is a must. For example, in many countries, law enforcement can lawfully access stored data or intercept communications under certain conditions. The important law enforcement tools necessary to carry out these exercises could be hampered by the use of uncontrolled cryptography, which may prevent lawful access to either plaintext or cryptographic keys of encrypted data. The privacy and integrity of data on communications systems are of vital importance. This paper addresses the issue at stake in communication security and the user's right to information on legal and social ramifications. This work describes how vital security techniques may be to information technology especially in the Internet era and how there could be a balance to individual privacy [Computers System and Network Security: Principles and Practice, 1999] and public safety in communication. This is one of the most essential issues at stake in communication security. It evaluates the threat caused by intrusion/violations of privacy rights by law enforcement agents and presents a better strategy of how crime can be traced and how criminals might be arrested by law enforcement agents without violating users' privacy rights. We propose an idea called SPLC – solution to privacy and lawful access conflict. SPLC realizes/instills online users' confidence and makes the Internet a fraud-free environment for e-commerce and other online activity.