Automated circuit sizing using Artificial Intelligence is a rapidly increasing area of interest, primarily thanks to its potential to accelerate product time-to-market and enhance employee satisfaction. A host of methods, rooted in different fundamental research philosophies, have been devised for this class of problems. While some of them perform well in terms of convergence speed , robustness has generally been given less attention. In this study we propose a novel automatic circuit sizing framework called Evolutionary Bayesian Optimization (EBO). It is a hybrid method combining the strengths of evolutionary computation techniques and Bayesian Optimization. EBO takes full advantage of parallel simulation infrastructure, by inherently using large batches of simulations. Our method is especially designed for multi-objective problems. Thus, it can optimize a large variety of circuits without the need of constructing figure of merit functions. Moreover, the strong emphasis on exploring the high-dimensional space of design variables ensures that EBO is robust and reliable across varying levels of problem complexity. We compare our framework with two state-of-the-art methods having different underlying philosophies and with arguably the most promising multi-objective evolutionary algorithm for this class of problems on four circuits: two proprietary voltage regulators, an open-source voltage regulator, and an open-source operational amplifier . The results show that EBO is superior to the other considered methods with regard to convergence speed and robustness. Generally, it can save between 30% and 70% circuit simulations compared to the next best performing method. Furthermore, EBO is the only method that finds circuit configurations that meet the specifications for all the considered circuits.

This article introduces a novel approach for building a robust Automatic Gesture Recognition system based on Surface Electromyographic (sEMG) signals, acquired at the forearm level. Our main contribution is to propose new constrained learning strategies that ensure robustness against adversarial perturbations by controlling the Lipschitz constant of the classifier. We focus on nonnegative neural networks for which accurate Lipschitz bounds can be derived, and we propose different spectral norm constraints offering robustness guarantees from a theoretical viewpoint. Experimental results on four publicly available datasets highlight that a good tradeoff in terms of accuracy and performance is achieved. We then demonstrate the robustness of our models, compared with standard trained classifiers in four scenarios, considering both white-box and black-box attacks.

There is a consistent dependence between integrated circuits (ICs) performance parameters and manufacturing process variations and capturing it at an early development phase represents a major ongoing topic in the semiconductor industry. Typically, this is addressed by the means of Monte Carlo (MC) simulations, where the device model parameters are randomly instantiated according to the technology variations based on a predefined nominal process. Thus, the resulted simulation data can only capture the effect of these variations. This offers little or no insight on the performance’s sensitivities to specific process variations or on the effect of altered statistical technology properties, as it may be the case of process drift or fab-to-fab migration. This article proposes a methodology for modeling the dependency of the device performances (i.e., electrical parameters—EPs) with the influential technology parameters (i.e., process control monitor parameters—PCMs), at an early stage (preSilicon). Using a set of standard MC co-simulations of PCM structures and the circuit schematics (to maintain consistent process variation), it employs a feature selection step to choose the influential PCMs and it trains a machine learning regression algorithm. Both are wrapped up in a Bayesian optimization (BO) framework to find the optimal feature set and the regression hyperparameters. The obtained regression model can explain the functional dependency of the EP on the influential PCMs. Thus, it directly enables sensitivity analysis (SA) to process variation and parametric yield prediction of the IC, as it will be illustrated for the case of an experimental Infineon Technologies product.

Self-training is a simple and efficient way of leveraging un-labeled speech data: (i) start with a seed system trained on transcribed speech; (ii) pass the unlabeled data through this seed system to automatically generate transcriptions; (iii) en-large the initial dataset with the self-labeled data and retrain the speech recognition system. However, in order not to pol-lute the augmented dataset with incorrect transcriptions, an important intermediary step is to select those parts of the self-labeled data that are accurate. Several approaches have been proposed in the community, but most of the works address only a single method. In contrast, in this paper we inspect three distinct classes of data-filtering for self-training, leveraging: (i) confidence scores, (ii) multiple ASR hypotheses and (iii) approximate transcriptions. We evaluate these approaches from two perspectives: quantity vs. quality of the selected data and improvement of the seed ASR by including this data. The proposed methodology achieves state-of-the-art results on Romanian speech, obtaining 25% relative improvement over prior work. Among the three methods, approximate transcriptions bring the highest performance gain, even if they yield the least quantity of data.

This paper proposes a method for speeding-up the verification process of integrated circuits, featuring waveform clustering of circuit response signals. The main objective is to automatically separate the signals into distinct groups that potentially exhibit visual similarities in order to aid the visual inspection/verification. As a first step, the proposed method extracts SIFT-like features by finding stable points of the signal over the scale space and computing robust descriptors able to describe their neighborhood. The resulted descriptors are quantized in order to be used in the clustering process as bag-of-words histograms. We demonstrate the validity of our method on a circuit waveform database containing several thousands of signals belonging to ten electrical tests.

This work proposes a new learning strategy for training a feedforward neural network subject to spectral norm and nonnegativity constraints. Our primary goal is to control the Lipschitz constant of the network in order to make it robust against adversarial perturbations of its inputs. We propose a stochastic projected gradient descent algorithm which allows us to adjust this constant in the training process. The algorithm is evaluated in the context of designing a fully connected network for Automatic Gesture Recognition based on EMG signals. We perform a comparison with the same architecture trained either in a standard manner or with simpler constraints. The obtained results highlight that a good accuracy-robustness trade-off can be achieved.

Although many efforts have been made in the last decade to enhance the speech and language resources for Romanian, this language is still considered under-resourced. While for many other languages there are large speech corpora available for research and commercial applications, for Romanian language the largest publicly available corpus to date comprises less than 50 hours of speech. In this context, Speech and Dialogue research group releases Read Speech Corpus (RSC) – a Romanian speech corpus developed in-house, comprising 100 hours of speech recordings from 164 different speakers. The paper describes the development of the corpus and presents baseline automatic speech recognition (ASR) results using state-of-the-art ASR technology: Kaldi speech recognition toolkit.

This paper proposes a method for explanation of the lifetime variation of power devices using data from different test stages. Understanding the lifetime variation is very useful in qualification, as well as in the characterization process, in order to improve the robustness of the power devices or to estimate more accurately the minimum guaranteed lifetime. Moreover, it helps design engineers better understand the root causes of the lifetime variation and use this knowledge to improve the performances of new power devices. In the proposed methodology, the variation of the lifetime is explained by the electrical parameters, measured before the stress-test. The Sensitivity Analysis presented here has the advantage of being simple and fast. It can be applied even when the number of test-runs is less than the number of factors. Moreover, it reveals not only linear correlations, but also quadratic effects and 2nd and 3rd order interactions. Eventually, the method provides the top of the most relevant electrical parameters which explain the lifetime variation. The validation of this approach has shown that 72% of the lifetime variation can be explained by the initial values of 5 electrical parameters.

The paper proposes a methodology for lifetime estimation of power devices at given applications operating conditions. The active cycling of power devices requires huge testing-time, because the process cannot be accelerated. For this reason, most often, the manufacturers provide information about the lifetime of power devices only for a few specific operating conditions. Most of the current methods are based on the junction temperature swing, which is very difficult to be measured or estimated. Instead, we propose an approach that, based on a few measurements of lifetime at given ambient temperatures, load currents and repetitive energies, is able to make lifetime prediction at any other set of operating conditions. The validation of the method was done by performing lifetime predictions in other operating conditions than those used for fitting the prediction function (metamodel) and it has shown a maximum relative error of 20%. With the proposed methodology, lifetime estimations of power devices can be made in the space of applications operating conditions, using optimal testing resources.

This study investigates the use of machine translated text for ASR domain adaptation. The proposed methodology is applicable when domain-specific data is available in language X only, whereas the goal is to develop a domain-specific system in language Y. Two semi-supervised methods are introduced and compared with a fully unsupervised approach, which represents the baseline. While both unsupervised and semi-supervised approaches allow to quickly develop an accurate domain-specific ASR system, the semi-supervised approaches overpass the unsupervised one by 10% to 29% relative, depending on the amount of human post-processed data available. An in-depth analysis, to explain how the machine translated text improves the performance of the domain-specific ASR, is also given at the end of this paper.

A new group-delay based spectral domain is explored for representation of speech signals and for extraction of robust features. The spectrum is computed using the group-delay functions defined on the autocorrelation of a short segment of speech. The features derived from this spectrum are easy to compute and are robust to the background noise. The invariance of the spectral shape to noise in this domain is demonstrated by comparing the group-delay spectrum to the Discrete Fourier transform (DFT) based spectrum and the LPC-derived spectrum. The new domain representation can be applied for parameter estimation as well as speech recognition. In this paper we present preliminary results of using such features in Speaker-Dependent (SD) as well Speaker-Independent (SI) recognition systems.