Exploiting the huge amount of data collected by industries is definitely one main challenges so-called Big Data era. In this sense, Machine Learning has gained growing attention in scientific community, as it allows to extract valuable information means statistical predictive models trained on historical process data. Semiconductor Manufacturing, most extensively employed data-driven applications Virtual Metrology, where a costly or unmeasurable variable estimated cheap and easy obtain...

The rise of industry 4.0 and data-intensive manufacturing makes advanced process control (APC) applications more relevant than ever for process/production optimization, related costs reduction, increased efficiency. One the most important APC technologies is virtual metrology (VM). VM aims at exploiting information already available in process/system under exam, to estimate quantities that are costly or impossible measure. Machine learning (ML) approaches foremost choice design solutions. A...

In manufacturing industries, it is of fundamental importance to detect anomalies in production order meet the required quality goals and limit number defective products that are accidentally delivered customers. Nevertheless, monitoring systems currently employed typically very simple rely on a set univariate control charts fail capture multivariate complex nature real-world industrial systems. such context, Machine Learning (ML)-based approaches for Anomaly Detection (AD) have proven be...

Virtual Metrology is one of the most prominent Advanced Process Control applications in Semiconductor Manufacturing. The goal to provide estimations quantities that are important for production and assess process quality, but costly or impossible be measured. solutions based on Machine Learning approaches. bottleneck developing generally feature extraction phase can time-consuming, deeply affect estimation performance. In particular, presence data with additional dimensions, such as time,...

In modern manufacturing scenarios, detecting anomalies in production systems is pivotal to keep high-quality standards and reduce costs. Even the Industry 4.0 context, real-world monitoring are often simple based on use of multiple univariate control charts. Data-driven technologies offer a whole range tools perform multivariate data analysis that allow implement more effective procedures. However, when dealing with complex data, common data-driven methods cannot be directly used, feature...

Metrology, which plays an important role in ensuring production quality modern manufacturing industries, incurs substantial costs both terms of the infrastructure required and time needed to perform measurements. In particular, semiconductor industry, measuring fundamental quantities on different sites a wafer surface is associated with increased time. To increase metrology efficiency, typical strategy limit number measured exploit statistical models (soft sensing) reconstruct profile....

Recent results show that features of adversarially trained networks for classification, in addition to being robust, enable desirable properties such as invertibility. The latter property may seem counter-intuitive it is widely accepted by the community classification models should only capture minimal information (features) required task. Motivated this discrepancy, we investigate dual relationship between Adversarial Training and Information Theory. We can improve linear transferability...

Embedding household appliances with smart capabilities is becoming common practice among major fabric-care producers that seek competitiveness on the market by providing more efficient and easy-to-use products. In Vertical Axis Washing Machines (VA-WM), knowing laundry composition fundamental to setting washing cycle properly positive impact both energy/water consumption performance. An indication of load typology (cotton, silk, etc.) typically provided user through a physical selector that,...

Deep Learning approaches have revolutionized in the past decade field of Computer Vision and, as a consequence, they are having major impact Industry 4.0 applications like automatic defect classification. Nevertheless, additional data, beside image/video itself, is typically never exploited classification module: this aspect, given abundance data data-intensive manufacturing environments (like semiconductor manufacturing) represents missed opportunity. In work we present use case related to...

In semiconductor manufacturing, metrology is generally a high cost, non-value added operation that impacts significantly on cycle time. As such, reducing wafer continues to be major target in manufacturing efficiency initiatives. Data-driven spatial dynamic sampling methodologies are here compared. Such strategies aim at minimizing the number of sites need measured across surface while maintaining an acceptable level profile reconstruction accuracy. The Spatial Dynamic Sampling approaches...

In laundry treatment appliances, the weight of loaded by user inside drum dramatically affects operating behavior. Therefore, it is important to obtain a good estimate said quantity in order correctly configure machine before washing/drying starts. Vertical Axis Washing Machines computed exploiting water absorbed clothes. However, such approach does not grant accurate results because absorption depends on clothes fabric. For this reason, we propose Soft Sensing for estimation that exploits...

In recent years, Data intensive technologies have become widespread in semiconductor manufacturing. particular, Virtual Metrology (VM) solutions had proliferated for quality, control and sampling optimization purposes. VM provide estimations of costly measures from already available data, allowing cost reduction increased throughput. While most the literature is focused on providing accurate methodological approach terms prediction accuracy, no work has previously investigated which are...

Wafer metrology is an expensive and time consuming activity in semiconductor manufacturing, but essential to support advanced process control, predictive maintenance other quality assurance functions. Keeping a minimum therefore desirable. In the context of spatial sampling wafers this has motivated development number data driven methodologies for optimizing wafer plans. Two such are considered paper. The first combines Principal Component Analysis Minimum Variance Estimation (PCA-MVE)...

Machine Learning-based Anomaly Detection approaches are efficient tools to monitor complex processes. One of the advantages such is that they provide a unique anomaly indicator, quantitative index captures degree 'outlierness' process at hand considering possibly hundreds or more variables same time, typical scenario in semiconductor manufacturing. drawbacks Root Cause Analysis not guided by system itself. In this work, we show effectiveness method, called DIFFI, equip Isolation Forest, one...

Various unsupervised greedy selection methods have been proposed as computationally tractable approximations to the NP-hard subset problem. These rely on sequentially selecting variables that best improve performance with respect a criterion. Theoretical results exist provide bounds and enable 'lazy greedy' efficient implementations for criteria satisfy diminishing returns property known submodularity. Recently, authors introduced Forward Selection Component Analysis (FSCA) which uses...

In this paper, we discuss how a promising word vector representation based on Probabilistic Word Embeddings (PWE) can be applied to Neural Information Retrieval (NeuIR). We illustrate PWE pros for text retrieval, and identify the core issues which prevent full exploitation of their potential. particular, focus application elliptical probabilistic embeddings, type PWE, NeuIR system (i.e., MatchPyramid). The main contributions paper are: (i) an analysis cons in NeuIR; (ii) in-depth comparison...

Deep Neural networks have gained lots of attention in recent years thanks to the breakthroughs obtained field Computer Vision. However, despite their popularity, it has been shown that they provide limited robustness predictions. In particular, is possible synthesise small adversarial perturbations imperceptibly modify a correctly classified input data, making network confidently misclassify it. This led plethora different methods try improve or detect presence these perturbations. this...

This paper introduces two simple techniques to improve off-policy Reinforcement Learning (RL) algorithms. First, we formulate RL as a stochastic proximal point iteration. The target network plays the role of variable optimization and value computes operator. Second, exploits functions commonly employed in state-of-the-art algorithms provide an improved action estimate through bootstrapping with limited increase computational resources. Further, demonstrate significant performance improvement...