Search results

The purpose of this paper is to describe a new approach to sentence representation learning leading to text classification using Bidirectional Encoder Representations from Transformers (BERT) embeddings. This work proposes a novel BERT-convolutional neural network (CNN)-based model for sentence representation learning and text classification. The proposed model can be used by industries that work in the area of classification of similarity scores between the texts and sentiments and opinion analysis.

The approach developed is based on the use of the BERT model to provide distinct features from its transformer encoder layers to the CNNs to achieve multi-layer feature fusion. To achieve multi-layer feature fusion, the distinct feature vectors of the last three layers of the BERT are passed to three separate CNN layers to generate a rich feature representation that can be used for extracting the keywords in the sentences. For sentence representation learning and text classification, the proposed model is trained and tested on the Stanford Sentiment Treebank-2 (SST-2) data set for sentiment analysis and the Quora Question Pair (QQP) data set for sentence classification. To obtain benchmark results, a selective training approach has been applied with the proposed model.

On the SST-2 data set, the proposed model achieved an accuracy of 92.90%, whereas, on the QQP data set, it achieved an accuracy of 91.51%. For other evaluation metrics such as precision, recall and F1 Score, the results obtained are overwhelming. The results with the proposed model are 1.17%–1.2% better as compared to the original BERT model on the SST-2 and QQP data sets.

The novelty of the proposed model lies in the multi-layer feature fusion between the last three layers of the BERT model with CNN layers and the selective training approach based on gated pruning to achieve benchmark results.

To ensure product quality within a manufacturing process, inspection processes are indispensable. One task of inspection planning is the selection of inspection characteristics. For optimization of costs and benefits, key characteristics can be defined by which the product quality can be checked with sufficient accuracy. The manual selection of key characteristics requires substantial planning effort and becomes uneconomic if many product variants prevail. This paper, therefore, aims to show a method for the efficient determination of key characteristics.

The authors present a novel Algorithm for the Selection of Key Characteristics (ASKC) based on an auto-encoder and a risk analysis. Given historical measurement data and tolerances, the algorithm clusters characteristics with redundant information and selects key characteristics based on a risk assessment. The authors compare ASKC with the algorithm Principal Feature Analysis (PFA) using artificial and historical measurement data.

The authors find that ASKC delivers superior results than PFA. Findings show that the algorithms enable the cost-efficient selection of key characteristics while maintaining the informative value of the inspection concerning the quality.

This paper fills an identified gap for simplified inspection planning with the method for the efficient selection of key features via ASKC.

This paper presents a micrograting‐based force sensor integrated with a surface micromachined silicon‐nitride probe suitable for characterizing microsurgery force on a single cell or embryo. The probe is supported by springs of a known spring constant, and the surgical penetration force is determined from displacement measurements. The optical‐encoder force sensor exhibits configurable sensitivity and dynamic range, allowing monitoring over a wide range of forces. The periodicity of the encoder response can be used for calibration of the injector displacement and to obtain information about the localized elastic properties of the target. We used a force sensor with a measured spring constant of 1.85 N/m for penetration force measurements on Drosophila embryos, and found a penetration force of 52.5 μN (±13.2 percent) and a membrane displacement of 58 μm (±5.2 percent).

Details of Electrical and Electronic Apparatus with Applications in the Maintenance and Operation of Aircraft, Missiles and Space Vehicles. These optical incremental encoders have been designed to optimize three basic characteristics ‐ very low price, high resolution, and dependability: low price through design simplifications, high volume production and maximum utilization of high impact moulded plastic components, high resolution (with equivalent accuracy), by computer‐designed optics and in house production and control of discs and masks; dependability, through careful material selection and the application of over 12 years' experience in the design and production of tens of thousands of encoders of a wide variety of types. A wide selection of resolutions can be provided. A binary (powers of two) and a decimal (multiples of 10) series are standard. The units presently being manufactured include a binary 2048 count per turn and a decimal 2000 count per turn. Earlier this year 500, 512, 1000 and 1024 count units were added to the line. Actually any number of direct output counts per turn up to a maximum of 2048 can be obtained on special order and counts as high as 4096 can be produced by electronic doubling of the basic count. This technique requires the use of the quadrature outputs normally furnished in encoders having direction sensing. Outputs from the encoder consist of a counting channel (A), a second counting channel (Q) in quadrature with the (A) channel, and an index marker channel (1) which provides a single pulse per revolution. Encoders arc supplied with any or all of these outputs cither at a low level direct from the photocells or with the outputs amplified and squared by internal electronics.

This paper aims to reduce the worst-case crosstalk effects for resistance, inductance and capacitance (RLC) interconnects using the bus encoding technique. In current nanoscale technology, power dissipation, propagation delay and crosstalk performance of interconnects determine the overall performance of a chip. Signal integrity issues due to crosstalk in the form of voltage glitches, overshoots, undershoots, undesirable noise, propagation speed ups and downs, etc. are some of the major deterrents for high-performance RLC modelled (VLSI) interconnects. This research paper primarily proposes two novel encoding methods (I and II) for RLC modelled interconnects to reduce the effect of crosstalk, simultaneous switching noise (SSN) and power consumption.

The proposed methods are based on the bus encoding method that is effective and well-suited for the reduction of the crosstalk noise. This method encodes or transforms incoming data in a manner that encoded data contain minimum or no crosstalk effects. The proposed encoding method uses the bus invert (BI) method. The proposed encoding methods are able to avoid the worst-case crosstalks while consuming lesser power during transmission in VLSI interconnects.

It is observed that the proposed encoders reduced/eliminated the worst-case crosstalk by reducing SSN. The encoding method I also reduces Type 0 crosstalk by 100 per cent, while Type 1 crosstalk is reduced by 36.4 per cent and Type 2 is reduced by 16.8 per cent. The average simultaneous switching is reduced by 51.1 per cent. Similarly, encoding method II reduces switching activity by 10.3 per cent, whereas the coupling activity is reduced by 35.4 per cent. Furthermore, encoding method II also reduced Type 0, Type 1 and Type 2 crosstalk by 100, 36.9 and 27.1 per cent, respectively. Hence, the proposed encoding methods reduced the worst-case crosstalk completely.

In VLSI technology, the reduction in feature size and the increase in operating frequency are quite rapid. This leads to higher propagation delay, crosstalk and power dissipation through the interconnects. Most of the previously proposed encoders/decoders have turned out to be unsuitable for RLC modelled interconnects. Hence, the proposed encoder would be extremely useful for crosstalk reduction in newer operating conditions.

The encoding method I identifies the harsh crosstalks, that is Type 0 and Type 1, in the inverted and non-inverted forms of incoming data with respect to the previous data. The data having minimum crosstalk in the inverted and non-inverted forms are only sent through the transmission line. The encoding method I also removes the worst-case crosstalk and simultaneously reduces other mild crosstalks. The removal of worst-case crosstalk improves the overall performance of the interconnect. The encoding method II identifies Type 2 crosstalk along with Type 0 and Type 1 similar to encoding method I. Furthermore, the encoding method II exhibits an improvement over method I in terms of reduction in crosstalk and power dissipation.

This paper proposes a novel encoding method to reduce worst-case crosstalk effects that reduces SSN. The proposed encoding methods achieve their purpose of crosstalk reduction for several technology nodes.

Review of the advancements in non-contact laser sensors for much more accurate real time speed as well as length measurements. The paper aims to discuss these issues.

In-depth review of a number of applications of laser sensors.

A wide range of manufacturers in the converting industries have solved measurement problems and realized major improvements in product quality, cost of production and productivity by applying laser measurement as well as reduction in scrap.

Others with the need to make very accurate measurements of speed and length in real time may find that laser sensors may provide the answer.

An expert insight into how to solve real time speed and length measurement problems.

Modern automation and process control demands high‐quality sensors that are sensitive to increasingly small rotations. Consequently, optical incremental encoders are finding an increasing number of applications for measuring position and rotational speed in mechanical engineering. As sensitivity increases, so they become more susceptible to electromagnetic interference (EMI).

This paper aim to solve the problem of low assembly success rate for 3c assembly lines designed based on classical control algorithms due to inevitable random disturbances and other factors,by incorporating intelligent algorithms into the assembly line, the assembly process can be extended to uncertain assembly scenarios.

This work proposes a reinforcement learning framework based on digital twins. First, the authors used Unity3D to build a simulation environment that matches the real scene and achieved data synchronization between the real environment and the simulation environment through the robot operating system. Then, the authors trained the reinforcement learning model in the simulation environment. Finally, by creating a digital twin environment, the authors transferred the skill learned from the simulation to the real environment and achieved stable algorithm deployment in real-world scenarios.

In this work, the authors have completed the transfer of skill-learning algorithms from virtual to real environments by establishing a digital twin environment. On the one hand, the experiment proves the progressiveness of the algorithm and the feasibility of the application of digital twins in reinforcement learning transfer. On the other hand, the experimental results also provide reference for the application of digital twins in 3C assembly scenarios.

In this work, the authors designed a new encoder structure in the simulation environment to encode image information, which improved the model’s perception of the environment. At the same time, the authors used the fixed strategy combined with the reinforcement learning strategy to learn skills, which improved the rate of convergence and stability of skills learning. Finally, the authors transferred the learned skills to the physical platform through digital twin technology and realized the safe operation of the flexible printed circuit assembly task.