Smart Data Solution Center Baden-Wuerttemberg

Jan 1, 2019

Homepage of SDSC project

Smart Data Solution Center (SDSC) a dynamic initiative aimed at propelling the digital transformation journey of small and medium-sized enterprises (SMEs).

Within the SDSC, I had the privilege of collaborating with two leading entities: the esteemed Karlsruhe Institute of Technology, providing robust computing power and technical expertise, and SICOS GmbH, entrusted with customer liaison and the ongoing maintenance and evolution of our digital platform.

My role as a data analyst from 2019 to 2024 was diverse and impactful, encompassing three primary domains: Data Analysis: I spearheaded projects offering tailored data modeling and analytical services to SMEs. Through meticulous examination of client data, we unearthed latent potentials or crafted bespoke models to address specific client needs. Each project spanned 2 to 3 months, fostering profound insights and actionable strategies.

Workshops: A cornerstone of our endeavor, workshops served as vibrant knowledge-sharing forums. We organized large-scale seminars featuring esteemed professors and seasoned entrepreneurs, delving into the nuances of digitalization, common challenges, and practical insights. Additionally, tailored workshops catered directly to the needs of individual SMEs, fostering collaborative problem-solving and strategic brainstorming sessions.

Website Development and Maintenance: Our digital footprint played a pivotal role in disseminating insights and fostering community engagement. I contributed to curating and maintaining a dynamic website replete with informative blogs, insightful demos, and invaluable resources pertinent to enterprise digitization. This platform served as a beacon of knowledge and a hub for fostering ongoing dialogue.

My journey within the SDSC has been enriching, marked by continuous learning, innovation, and meaningful collaborations. Through my multifaceted involvement across data analysis, workshops, and digital platform management, I’ve been privileged to contribute to the advancement of SMEs on their transformative digital journeys.

Publications

ExTea: An Evolutionary Algorithm-Based Approach for Enhancing Explainability in Time-Series Models

We introduce an EXplainable artificial intelligence method targeting Time-series model based on Evolutionary Algorithm (ExTea). ExTea conceptualizes explanations as evolving individuals and employs an innovative pyramidal structure for optimizing potential explanations, categorized into newborn, tested, and elite stages.

黄逸然, Yexu Zhou, Haibin Zhao, Likun Fang, Till Riedel, Michael Beigl

ExTea: An Evolutionary Algorithm-Based Approach for Enhancing Explainability in Time-Series Models

Generate Explanations for Time-series classification by ChatGPT

In this paper, we aim to introduce a novel method for explaining time-series classification, leveraging the capabilities of ChatGPT to enhance the interpretability of results and foster a deeper understanding of feature contributions within time-series data.

Zhechang Xue, 黄逸然, Hongnan Ma, Michael Beigl

A Survey on Wearable Human Activity Recognition: Innovative Pipeline Development for Enhanced Research and Practice

we review WHAR articles from 2021 to 2023 and introduce an innovative WHAR pipeline, emphasizing a research-focused approach. This new pipeline offers distinct advantages; it provides researchers with a clear and systematic categorization of WHAR articles, thereby enhancing understanding of the field. For practitioners, it facilitates the selection of customized methods for each stage, thereby optimizing final assembled model efficacy.

黄逸然, Yexu Zhou, Haibin Zhao, Till Riedel, Michael Beigl

Enhancing Efficiency in HAR Models: NAS Meets Pruning

Real-time monitoring of human activities using wearable devices often requires the deployment of machine learning models on resource-constrained edge devices. Stateof- the-art Human Activity Recognition neural network models, while effective, already suffer from excessive size and complexity. Furthermore, our systematic analysis reveals that, even worse, the computational cost and model size of most SOTA HAR models escalate significantly with an increase in sensor channels. In this work, we propose an integrated neural architecture search framework to further lighten HAR models. To counteract the negative effects of scaling at deployment time, the proposed framework simultaneously selects and reduces the number of sensor channels, prunes filters, and decreases the temporal dimensions while training the model on optimized hardware.

Yexu Zhou, Tobias King, 黄逸然, Likun Fang, Tobias Röddiger, Till Riedel, Michael Beigl

Enhancing Efficiency in HAR Models: NAS Meets Pruning

Optimizing AutoML for Tiny Edge Systems: A Baldwin-effect Inspired Genetic Algorithm

Tiny edge systems used in IoT devices, wearables or smart textiles are characterized by the need of processing complex sensor data streams under various device constraints. Due to the high number of constraints and the complexity of the optimization of the hyper-parameter space for machine learning based processing, genetic algorithms (GAs) seem to be a perfect fit to enable AutoML for those embedded devices. We introduce a novel Genetic Algorithm (GA) customized for AutoML tasks, addressing the unique challenges posed by highly embedded machine learning domains.

黄逸然, Yexu Zhou, Haibin Zhao, Till Riedel, Michael Beigl

Optimizing AutoML for Tiny Edge Systems: A Baldwin-effect Inspired Genetic Algorithm

A Survey on Wearable Human Activity Recognition: Innovative Pipeline Development for Enhanced Research and Practice

We review WHAR articles from 2021 to 2023 and introduce an innovative WHAR pipeline, emphasizing a research-focused approach. This new pipeline offers distinct advantages. it provides researchers with a clear and systematic categorization of WHAR articles, thereby enhancing understanding of the field. For practitioners, it facilitates the selection of customized methods for each stage, thereby optimizing final assembled model efficacy.

黄逸然, Yexu Zhou, Haibin Zhao, Till Riedel, Michael Beigl

Investigating Passive Haptic Learning of Piano Songs Using Three Tactile Sensations of Vibration, Stroking and Tapping

Passive Haptic Learning (PHL) is a method by which users are able to learn motor skills without paying active attention. In past research, vibration is widely applied in PHL as the signal delivered on the participant’s skin. We investigated whether different tactile sensations (tapping, light stroking, and vibration) influence the learning effect of PHL in this work. We built three wearable systems corresponding to the three sensations respectively.

Fang Likun, 黄逸然, M{\"u}ller Timo, Pescara Erik, Fischer Nikola, Michael Beigl

McXai: Local Model-Agnostic Explanation As Two Games

To this day, various approaches for providing local explanation of black box machine learning models have been introduced. Despite these efforts, existing methods suffer from deficiencies such as being difficult to comprehend, only considering one feature at a time and disregarding inter-feature dependencies, lacking meaningful values for each feature, or only highlighting features that support the model’s decision. To overcome these drawbacks, this study presents a new approach to explain the predictions of any black box classifier, called Monte Carlo tree search for eXplainable Artificial Intelligence (McXai).

黄逸然, Nicole Schaal, Michael Hefenbrock, Yexu Zhou, Till Riedel, Michael Beigl

McXai: Local Model-Agnostic Explanation As Two Games

randomHAR: Improving Ensemble Deep Learners for Human Activity Recognition with Sensor Selection and Reinforcement Learning

Deep learning has proven to be an effective approach in the field of Human activity recognition (HAR), outperforming other architectures that require manual feature engineering. Despite recent advancements, challenges inherent to HAR data, such as noisy data, intra-class variability and inter-class similarity, remain. To address these challenges, we propose an ensemble method, called randomHAR.

黄逸然, Yexu Zhou, Likun Fang, Till Riedel, Michael Beigl

randomHAR: Improving Ensemble Deep Learners for Human Activity Recognition with Sensor Selection and Reinforcement Learning

Standardizing Your Training Process for Human Activity Recognition Models: A Comprehensive Review in the Tunable Factors

In recent years, deep learning has emerged as a potent tool across a multitude of domains, leading to a surge in research pertaining to its application in the wearable human activity recognition (WHAR) domain. Despite the rapid development, concerns have been raised about the lack of standardization and consistency in the procedures used for experimental model training, which may affect the reproducibility and reliability of research results. In this paper, we provide an exhaustive review of contemporary deep learning research in the field of WHAR and collate information pertaining to the training procedure employed in various studies.

黄逸然, Haibin Zhao, Yexu Zhou, Till Riedel, Michael Beigl

Standardizing Your Training Process for Human Activity Recognition Models: A Comprehensive Review in the Tunable Factors

State Graph Based Explanation Approach for Black-Box Time Series Model

In recent years, there has been a growing trend in the utilization of Artificial Intelligence (AI) technology to construct human-centered systems that are based on implicit time series information, ranging from contextual recommendations on smartwatches to human activity recognition on production workshop. Despite the advantages of these systems, the opaqueness and unpredictability of these systems for users has elicited concerns. In this paper, we propose a novel explanation method named State-graph Based eXplanable Artificial Intelligent (SBXAI), which exhibits the sequential relationship between time periods through directed circular graphs while emphasizing the importance of each time period in an instance.

黄逸然, Chaofan Li, Hansen Lu, Till Riedel, Michael Beigl

State Graph Based Explanation Approach for Black-Box Time Series Model

Automatic Feature Engineering Through Monte Carlo Tree Search

The performance of machine learning models depends heavily on the feature space and feature engineering. Although neural networks have made significant progress in learning latent feature spaces from data, compositional feature engineering through nested feature transformations can reduce model complexity and can be particularly desirable for interpretability. To address these shortcomings, we propose a reinforcement learning-based automatic feature engineering method, which we call Monte Carlo tree search Automatic Feature Engineering (mCAFE).

黄逸然, Michael Hefenbrock, Yexu Zhou, Likun Fang, Till Riedel, Michael Beigl

DragTapVib: An On-Skin Electromagnetic Drag, Tap, and Vibration Actuator for Wearable Computing

The skin, as the largest organ distributed all over the human body, offers excellent opportunities for different kinds of input stimuli. However, most of the haptic devices can only render single sensations or they need to combine multiple complex components for generating multiple sensations. We present “DragTapVib” in this paper, a novel, ultra-low-cost, wearable actuator that can reliably provide dragging, tapping, and vibrating sensations to the user.

Likun Fang, 黄逸然, Ting Zhu, Pescara Erik, Yexu Zhou, Michael Beigl

DragTapVib: An On-Skin Electromagnetic Drag, Tap, and Vibration Actuator for Wearable Computing

Universal Distributional Decision-Based Black-Box Adversarial Attack with Reinforcement Learning

The vulnerability of the high-performance machine learning models implies a security risk in applications with real-world consequences. In this work, we propose a pixel-wise decision-based attack algorithm that finds a distribution of adversarial perturbation through a reinforcement learning algorithm.

黄逸然, Yexu Zhou, Michael Hefenbrock, Likun Fang, Till Riedel, Michael Beigl

Automatic remaining useful life estimation framework with embedded convolutional LSTM as the backbone

An essential task in predictive maintenance is the prediction of the Remaining Useful Life (RUL) through the analysis of multivariate time series. Using the sliding window method, Convolutional Neural Network (CNN) and conventional Recurrent Neural Network (RNN) approaches have produced impressive results on this matter, due to their ability to learn optimized features. To exploit the multi-level temporal information, many approaches are proposed which combine CNN and RNN models. In this work, we propose a new LSTM variant called embedded convolutional LSTM (ECLSTM).

Yexu Zhou, Michael Hefenbrock, 黄逸然, Till Riedel, Michael Beigl

NeuralIO: Indoor outdoor detection via multimodal sensor data fusion on smartphones

The Indoor Outdoor (IO) status of mobile devices is fundamental information for various smart city applications. In this paper we present NeuralIO, a neural network based method to deal with the Indoor Outdoor (IO) detection problem for smartphones. Multimodal data from various sensors on a smartphone are fused through neural network models to determine the IO status. A data set consisting of more than 1 million samples is constructed. We test the performance of an early fusion scheme in various settings. NeuralIO achieves above 98% accuracy in 10-fold cross-validation and above 90% accuracy in a real-world test.

Long Wang, Lennard Sommer, 黄逸然, Yexu Zhou, Till Riedel, Michael Beigl

Events

Presentation at ECMLPKDD 2024

Presentation for the conference paper - ExTea An Evolutionary Algorithm-Based Approach for Enhancing Explainability in Time-Series Models.

Sep 9, 2024 — Sep 13, 2024 VILNIUS, LITHUANIA

Presentation at ECMLPKDD 2024

Presentation at XAI 2024

Presentation for the conference paper - Generate Explanations for Time-series classification by ChatGPT.

Jul 17, 2024 — Jul 19, 2024 Valletta, Malta

Presentation at XAI 2024

Presentation at IJCNN 2024

Presentation for the conference paper - A Survey on Wearable Human Activity Recognition; Innovative Pipeline Development for Enhanced Research and Practice.

Jun 30, 2024 — Jul 5, 2024 YOKOHAMA, JAPAN

Presentation at Percom 2024

Presentation for the conference paper - Optimizing AutoML for Tiny Edge Systems, A Baldwin-effect Inspired Genetic Algorithm.

Mar 11, 2024 — Mar 15, 2024 Biarritz, France