Welcome to 2025, MCL family! Here’s to a year full of happiness, success, and the courage to chase your dreams. May this be a time of growth, discovery, and opportunities that take you to new heights. Let’s make it a year to remember—graceful, bold, and uniquely ours. Cheers to the journey ahead!
Photo credits: https://www.freepik.com/
As 2024 draws to a close, we reflect on a year of growth, success, and countless blessings at MCL. This year, we bid a fond farewell to our esteemed graduates, whose impactful research leaves an enduring legacy as they step into exciting new chapters. At the same time, we joyfully welcomed fresh faces into our MCL family—brimming with energy and enthusiasm for groundbreaking research.
Through dedication, collaboration, and resilience, our members have achieved remarkable milestones, including publishing exceptional work in prestigious journals and conferences. These accomplishments are a testament to the passion and hard work of every individual in our community.
Now, as the holiday season unfolds with its sparkle of joy and the warmth of togetherness, we take a moment to celebrate the achievements of this past year and the unity that binds us as the MCL family.
From all of us at MCL, we wish you a Merry Christmas filled with cheer, love, and laughter. May this festive season bring you peace, joy, and countless reasons to smile!
Pathology imaging is a key method in medical diagnostics, offering detailed information on tissue structures and biomarkers. However, traditional approaches to measuring biomarker intensity can be time-consuming, which is why new machine-learning methods are needed to predict these biomarkers directly from images. By identifying and analyzing biomarkers in this way, doctors can better predict patient grades, postoperative responses, and overall disease progression. This approach also presents significant commercial potential.
To address this issue, we adopted the Green U-Shaped Learning (GUSL) pipeline. Specifically, we extracted 15 randomly selected regions of interest (ROIs) from each pathology slide and resized them into patches. Each ROI was then analyzed to determine its biomarker density. GUSL utilizes Green Learning techniques like PixelHop, RFT, and LNT to integrate features from different spatial scales and refine predictions level by level. While this method shows promising results, it requires further adjustments to effectively handle ROIs with larger spatial sizes.
Image segmentation is a computer vision process involving dividing an image into segments based on shared characteristics such as colour, texture, or intensity. Semantic segmentation, specifically, is a type of image segmentation assigning a class label to each pixel in an image. This approach is widely used in autonomous driving, object detection and medical imaging.
To address such problems, we have proposed a green image segmentation approach that identifies each image segment without requiring backpropagation. The image is resized and divided into non-overlapping patches, with each patch labelled as pure or impure based on its class composition. Impure patches are iteratively enlarged and subdivided until all patches are labelled by class. This is a promising method, but it still requires further refinement to deal with incorrectly classified large patches and object boundaries.
For over 20 years, the Thanksgiving Luncheon has been a cornerstone of MCL’s community spirit, offering a chance to celebrate gratitude and connection. This year, on November 28, 2024, the tradition continued as the MCL family gathered at Shiki Seafood Buffet to share a memorable meal.
The luncheon was more than just an opportunity to enjoy great food; it was a moment to step away from the daily grind and reconnect with friends and colleagues. The lively atmosphere was filled with laughter and meaningful conversations, reminding everyone of the strength and warmth of the MCL community.
Such an event wouldn’t have been possible without the guidance of Professor Kuo, whose dedication keeps this tradition alive, and the students who put in the effort to ensure everything ran smoothly.
As another Thanksgiving Luncheon joins the books, the MCL family looks forward to many more years of celebration, connection, and gratitude. Happy Thanksgiving to all!
Deep learning (DL) models have driven advancements in AI and machine learning but face challenges such as interpretability, susceptibility to adversarial attacks, dependency on pre-trained networks, and high computational demands. These limitations hinder their deployment on mobile and edge devices.
Chee-An Yu, inspired by the concept of Green AI/ML (GL) introduced by Kuo, focuses on developing energy-efficient, mathematically transparent models with small sizes and low complexity. These models excel in limited-data scenarios and are suitable for both cloud and edge environments. The current project explores GL methods to learn RF signatures for detecting jamming signals and reconstructing the original signal using the Green U-Shape Learning (GUSL) pipeline.[1-3]
Initial applications of GL in wireless communications have shown promise, with efficient performance in few-shot learning tasks and reduced computational requirements.
References:1. Chen Chung, C.-C. Jay Kuo, and Shang-Ho Tsai, “Effective and efficient beam tracking with green learning,” IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Valencia, Spain, September 2-5, 2024.
2. Tzu-Ching Liao, Wan-Jen Huang, and C.-C. Jay Kuo, “Green-learning based design of RIS-assisted MIMO systems based on implicit CSI,” IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Valencia, Spain, September 2-5, 2024.
3. Kai-Rey Liu, Sau-Hsuan Wu, C.-C. Jay Kuo, Lie-Liang Yang, and Kai-Teng Feng, “3D positioning via green learning in mmWave hybrid beamforming systems,” VTC 2024-Spring, Singapore, June 24-27, 2024.
MCL Director, Professor C.-C. Jay Kuo, received the Distinguished Alumni Award from his Alma Mater, National Taiwan University (NTU), at its 96 th Anniversary Ceremony on November 15 (Friday), 2024, in Taipei, Taiwan. The university was founded in 1928 during Japanese rule as the seventh of the Imperial Universities. The university comprises 11 colleges, 56 departments, 133 graduate institutes, and 60 research centers.
Professor Kuo studied as an undergraduate in the Electrical Engineering Department at NTU from 1976 to 1980. He said, “NTU provided an excellent environment for me to make good friends, explore new things, and build academic background, so I became more mature and independent. It was a memorable period of time in my life.” Professor Kuo received this honor for his contributions to multimedia technologies. He added, “I want to share this honor with all of my students and my family. Their love, trust, and efforts make this award possible. I am very grateful.”
Feature extraction in Computer Vision aims to pinpoint relevant information in images. While Convolutional Neural Networks (CNNs) implicitly learn feature importance, tools like Grad-CAM can help interpret which image regions influence predictions. More recently, Transformer-based models like ViT and DINO have gained traction by incorporating attention mechanisms that naturally focus on critical input parts, improving interpretability.
Building on these ideas, Jie-En Yao from MCL lab proposes a novel approach: Forward Green-Attention, which identifies essential regions in an image without requiring backpropagation. This method utilizes SHAP values from XGBoost to highlight regions that push model predictions positively or negatively. High positive SHAP values reveal areas driving positive classifications, while negative values indicate regions leading to negative classifications. Though promising, this approach is limited by the receptive field size and feature-dependent interpretability, highlighting areas for further refinement.
Word embedding is a fundamental task in natural language processing. It converts each word into a representation in a vector space. A challenge with word embedding is that, as the vocabulary grows, the vector space’s dimension increases – leading to a vast model size. Storing and processing word vectors are resource-demanding, especially for mobile edge-devices applications.
Jintang Xue, a PhD student at MCL, has proposed a dimension reduction method called WordFS [1] for pre-trained word embeddings. WordFS combines a post-processing algorithm (PPA) and weakly- supervised feature selection with limited word similarity pairs. It is simpler, more efficient, and more effective than existing approaches. Experimental results show it excels in word similarity tasks and generalizes well across downstream tasks. WordFS effectively reduces embedding dimensions with lower computational costs.
[1] Xue, Jintang, et al. “Word Embedding Dimension Reduction via Weakly-Supervised Feature Selection.” arXiv preprint arXiv:2407.12342 (2024).
We are so happy to welcome a new MCL member, Hong-En Chen joining MCL this semester. Here is a quick interview with Hong-En:
1. Could you briefly introduce yourself and your research interests?
My name is Hong-En Chen, a PhD student in Electrical Engineering. My research focuses on 3D object generation, aiming to integrate Green Learning for light-weight, interpretable 3D generation. Ultimately, I aim to develop intuitive, reliable generative applications that bring users’ imaginations to life.
2. What is your impression about MCL and USC?
Before coming to USC, I expected a smaller, more modern campus, but I was impressed by the uniform red-brick architecture and vibrant student life, with numerous events beyond academics. My first impression of MCL is that it has a large, dedicated team focused on Green Learning. After arriving, I’ve enjoyed working with the team, learning from their valuable experience, and collaborating on interpretable AI reserach.
3. What is your future expectation and plan in MCL?
In MCL, I aim to contribute not only to 3D generation but also to advancing the core library of Green Learning. I plan to build a strong mathematical foundation to design more interpretable and efficient models for future research. Additionally, I hope to deepen my relationships with team members, fostering collaboration and generating innovative ideas, while creating meaningful memories throughout my PhD journey.
