Selection of network and learning parameters for an adaptive neural robotic control scheme

This paper presents the results of a study in the design of a neural network based adaptive robotic control scheme. The neural network used here is a two hidden layer feedforward network and the learning scheme is the well-known backpropagation algorithm. The neural network essentially provides the inverse of the plant and acts in conjunction with a standard PD controller in the feedback loop. The objective of the controller is to accurately control the end position of a single link manipulator in the presence of large payload variations, variations in the link length and also variations in the damping constant. Based on results of this study, guidelines are presented in selecting the number of neurons in the hidden layers and also the parameters for the learning scheme used for training the network. Results also indicate that increasing the number of neurons in the hidden layer will improve the convergence speed of learning scheme up to a certain limit beyond which the addition of neurons will cause oscillations and instability. Guidelines for selecting the proper learning rate, momentum and fast backpropagation constant that ensure stability and convergence are presented. Also, a relationship between the r.m.s. error and the number of iterations used in training the neural network is established.     

Effects of flux enhancing polymer on the characteristics of sludge in membrane bioreactor process.

A newly developed membrane performance enhancer (MPE) was used to prevent membrane fouling in a membrane bioreactor (MBR) process. It transpired that 1,000 mg/l of MPE reduced polysaccharide levels from 41 mg/I to 21 mg/I on average under the experimental condition. Repeated experiments also confirmed that 50-1,000 mg/l of MPE could reduce membrane fouling significantly and increase the intervals between membrane cleanings. Depending on MPE dosages and experimental conditions, trans-membrane pressure (TMP) increase was suppressed for 20-30 days, while baseline TMP surged within a few days. In addition, MPE allowed MBR operation even at 50,000 mg/l of total solid and reduced permeate COD. However, no evidence of toxicity for sludge was found from respiratory works.     

Novel Short-chain Crosslinked Cationomeric Polyurethanes

Summary Novel PUs containing pyridinium moieties were synthesized by chain extending isocyanate endcapped prepolymers with N, N’-bis (2-hydroxyethyl) isonicotinamide. The pyridinium moieties in the PUs were chemically crosslinked using short-chain divalent quaternising agents. The polyurethane cationomers were characterized by spectral, thermal and mechanical analysis. Spectral results confirmed the quaternisation of tertiary nitrogen leading to crosslinking. Compared to conventional PUs, the crosslinked PU networks exhibited improved thermal stability. The damping value (i.e.) tan δ for cationomers were improved over a broad temperature range when compared to conventional PU.     

Effect of temperature on recombinant protein production using the Bm5/Bm5.NPV expression system

A series of experiments have been conducted using a recombinant baculovirus/insect cell expression system (Bm5/Bm5.NPV.CAT) to establish the optimum temperature for both cell growth and virus infection. Bm5 cell growth was found to be limited at temperatures below 22°C and ceased completely at temperatures above 34°C. In the range between 24 and 28°C, final cell densities always reached 96% of the highest achievable viable cell density. The shortest population doubling time was obtained at 28°C. Overall, a consistent increase in metabolism with increasing temperatures was observed. During the infection/viral replication phase, an increase in the temperature from 25 to 31°C resulted in a faster decrease in viable cell density and an earlier production of chloramphenicol acetyltransferase (CAT). Furthermore, protein yield at temperatures above 28°C was significantly reduced. Overall, the best temperature for the infection phase for the Bm5/Bm5.NPV expression system was found to be 25°C when the cells are cultured in serum free media.     

SHARED: An information model for cooperative product development

One fundamental issue in developing collaborative engineering systems is the representation of product information which supports communication and coordination. This product information includes not only the geometric and physical properties of the product and its parts, but also information about functions, constraints and the design rationale. In this paper, we describe an information model, SHARED, which was developed for encoding product information in DICE, a distributed and integrated environment for computer-aided engineering. SHARED provides multiple levels of both functional and geometric abstractions, multiple views and techniques for maintaining consistency between the various abstractions and views. These elements are essential for a good representation model of product information. The use of the SHARED model is illustrated through an example, depicting the various representations of a product as it evolves through the design process. The SHARED model has been implemented over a distributed OODBMS as a toolkit/framework for developing environments which need to model, manipulate and communicate product information between distributed cooperating applications, while supporting coordination between them.     

A comparison of statistical relational learning and graph neural networks for aggregate graph queries

Statistical relational learning (SRL) and graph neural networks (GNNs) are two powerful approaches for learning and inference over graphs. Typically, they are evaluated in terms of simple metrics such as accuracy over individual node labels. Complex aggregate graph queries (AGQ) involving multiple nodes, edges, and labels are common in the graph mining community and are used to estimate important network properties such as social cohesion and influence. While graph mining algorithms support AGQs, they typically do not take into account uncertainty, or when they do, make simplifying assumptions and do not build full probabilistic models. In this paper, we examine the performance of SRL and GNNs on AGQs over graphs with partially observed node labels. We show that, not surprisingly, inferring the unobserved node labels as a first step and then evaluating the queries on the fully observed graph can lead to sub-optimal estimates, and that a better approach is to compute these queries as an expectation under the joint distribution. We propose a sampling framework to tractably compute the expected values of AGQs. Motivated by the analysis of subgroup cohesion in social networks, we propose a suite of AGQs that estimate the community structure in graphs. In our empirical evaluation, we show that by estimating these queries as an expectation, SRL-based approaches yield up to a 50-fold reduction in average error when compared to existing GNN-based approaches.

     

AHW-BGOA-DNN: a novel deep learning model for epileptic seizure detection

Neural Computing and Applications - “Brain–Computer Interface” (BCI)—a real-life support system provides a way for epileptic patients to improve their quality of life. In...     

Multipartite entangled magnon states as quantum communication channels

We investigate the entanglement properties of the two magnon states and explicate conditions under which, the two magnon state becomes useful for several quantum communication protocols. We systematically study the temporal behaviour of concurrence to find out the effect of exchange interaction on entanglement. The two magnon state, which is potentially realizable in quantum dots using Heisenberg exchange interaction, is found to be suitable for carrying out deterministic teleportation of an arbitrary two qubit composite system. Further, conditions for which the channel capacity reaches “Holevo bound”, allowing four classical bits to be transmitted through two qubits are derived. Later, an unconventional protocol is given to demonstrate that this state can be used for sharing of a two qubit entangled state among two parties.     

CMOS integrated elliptic diaphragm capacitive pressure sensor in SiGe MEMS

A novel CMOS integrated Micro-Electro-Mechanical capacitive pressure sensor in SiGe MEMS (Silicon Germanium Micro-Electro-Mechanical System) process is designed and analyzed. Excellent mechanical stress–strain behavior of Polycrystalline Silicon Germanium (Poly-SiGe) is utilized effectively in this MEMS design to characterize the structure of the pressure sensor diaphragm element. The edge clamped elliptic structured diaphragm uses semi-major axis clamp springs to yield high sensitivity, wide dynamic range and good linearity. Integrated on-chip signal conditioning circuit in 0.18 μm TSMC CMOS process (forming the host substrate base for the SiGe MEMS) is also implemented to achieve a high overall gain of 102 dB for the MEMS sensor. A high sensitivity of 0.17 mV/hPa (@1.4 V supply), with a non linearity of around 1 % is achieved for the full scale range of applied pressure load. The diaphragm with a wide dynamic range of 100–1,000 hPa stacked on top of the CMOS circuitry, effectively reduces the combined sensor and conditioning implementation area of the intelligent sensor chip.