🚨 ML for Gas Sensors: Anomaly Detection & Domain-Aware Modeling

“Graphene-Based Sensor Data Analytics” is a first-of-its-kind international workshop that combines the nanomaterial science of graphene with the data-centric world of AI and IoT analytics. Graphene’s ultra-sensitivity makes it ideal for sensors in biomedicine, environmental monitoring, chemical detection, and flexible electronics—but leveraging these sensors at scale requires sophisticated data processing pipelines.

Participants will explore the fundamentals of graphene-based sensor mechanisms, and build practical workflows using tools like Python, pandas, scikit-learn, TensorFlow, MATLAB, and real-time signal analysis libraries to process, classify, and predict sensor responses.

• Bridge the gap between sensor hardware innovation and intelligent data use

• Promote cross-disciplinary learning between nanoscience and AI

• Empower participants to contribute to next-gen sensor networks and smart systems

• Foster innovation in sustainable, scalable, and real-time sensing platforms

• Prepare researchers to publish or commercialize sensor-based data solutions

Day 1: Signal Preprocessing for SAW Gas Sensors

• Introduction to SAW Gas Sensors

  • Working principles and applications

  • Common signal characteristics and noise sources

• Signal Preprocessing Techniques

  • Filtering: Low-pass, band-pass, median filters

  • Baseline correction and normalization

• Feature Extraction & Dimensionality Reduction

  • Time-domain and frequency-domain features

  • PCA, FFT, and wavelet transforms for SAW data

Day 2: Anomaly Detection Using Autoencoders

• Basics of Autoencoders

  • Architecture: Encoder, bottleneck, decoder

  • Training for reconstruction accuracy

• Autoencoders for Anomaly Detection

  • Loss-based detection of anomalous gas responses

  • Performance evaluation metrics (AUC, precision-recall)

• Implementation & Case Study

  • Building an autoencoder model in Python (Keras/PyTorch)

  • Real SAW data analysis with labeled anomalies

Day 3: Transfer Learning for New Analytes

• Introduction to Transfer Learning

  • What is transfer learning and why it matters

  • Types: Feature-based, fine-tuning, domain adaptation

• Applying Transfer Learning to Sensor Data

  • Transferring models across sensor types or analyte classes

  • Handling distribution shift and domain generalization

• Advanced Techniques & Future Directions

  • Meta-learning, few-shot learning, and continual learning

  • Preparing your model for deployment in dynamic environments

• Nanotechnology researchers and material scientists

• Sensor engineers and IoT hardware developers

• AI/ML professionals working in biomedical or environmental sensing

• Researchers in wearable and flexible electronics

• UG/PG/PhD students in physics, electronics, materials, or data science