Mao-Lin Li

Google Scholar: https://tinyurl.com/y7b9a8p5 • Matrix Factorization with Interval-Valued Data (TKDE 2019, ICDE 2020) - Proposed interval-valued matrix factorization techniques to achieve meaningful latent semantic analysis for interval-valued data. - Achieved 30% accuracy improvement in low-rank decomposition with applications include image classification/clustering and recommendation systems. • GTT: Guiding the Tensor Train in Selecting the Decomposition Sequence (SISAP 2020 Best paper candidate, ISJ 2022) - Proposed a data-driven approach for selecting Tensor-train decomposition order in high dimensional data. - Increased 20% selection quality without exhaustive searching in 15 real-world data sets. • W2FM: The Doubly Warped Factorization Machine (PAKDD 2020) - Proposed a method that leverages multiple space warping strategies to improve the representational ability of factorization machines (FMs). - Achieved 45% improvement than conventional FMs in accuracy with less inference time. - Integrated with a neural network architecture and achieves 17% improvement than modern competitors. • CTTD: Causally Informed Tensor Train Decomposition - Proposed a method that leverages the structural information in a given causal graph and recommends a suitable causally informed decomposition sequence for TT-decomposition. - Achieves 30% accuracy gain compares with expected accuracy. • Personalized PageRank in Uncertain Graphs with Mutually Exclusive Edges (SIGIR 2017) - Propose an efficient Uncertain Personalized PageRank (UPPR) approach on graphs with edge uncertainties. Utilized Sherman-Morrison Lemma to approximate matrix inversion efficiently in PPR computation. - Achieved 2 orders of performance improvement than traditional approaches with comparable accuracy.

• Developed end-to-end retrieval algorithms for boosting FB Reels user mimicry. - Developed a launch-ready generator based on two tower sparse network (TTSN) neural net model for boosting user mimicry with 4% improvement. - Designed a scored-based retrieval scenario that leverages user engagements o Created a daily data pipeline to aggregate user engagements, e.g., numbers of music clicks, effect clicks, hashtag clicks) as reference metrics. o Designed a model to combine user engagements to select candidates which prompt users to mimic videos. - Established a clustering method for retrieving trending Reels based on visual representation. o Created a data pipeline to aggregate trending hashtags and visual embedding for fetching candidates. o Applied K-means based method for clustering video candidates for video recommendation.

• Explored the effect of various feature meta-settings and model architecture settings on the feature selection results/model performance. • Achieved 15% to 17% performance gain with selected configuration compare to production models.

• Defined features from multi-modal, time-series data to represent several user properties and dynamic behaviors. • Designed a classification algorithm to identify residency within specific areas.

Developed a personal healthcare system that monitoring patients’ health status in real-time. • Implemented system with component-based approach in Java, including graphic users interface (GUI) in Java Swing library, server/client communication, remote database manipulation and multiple sensor controlling.

Teaching Assistant - Introduction to Computer Architecture, (Fall 2012) - Intermediate Programming Using JAVA, (Fall 2012, Summer 2013, Fall 2013) - Introduction to Computer Programming, (Summer 2013, Fall 2013) Fellowship - Arts & Science Graduate Fellow in Computer Science Department, Spring 2013

Maintaining conference database and registration system with PHP and HTML, equipments setting and configuration.

Teaching computer knowledge to the workers in processing industry.

• Designed a two-phase cycle count accurate (CCA) arbiter model that speed up simulation performance 20 times than traditional cycle accurate (CA) approach with 100% accuracy in Multi-Processor System on Chip. • Implemented in C++ with SystemC library. • Related publication won Outstanding Paper Award out of 72 papers in international workshop.

Master of Computer Science. • Designed a two-phase arbiter model that speed up simulation performance 20 times than traditional simulation approach in Multiprocessor System on Chip platform. • Implemented in C++ with SystemC library. • Related publication won Outstanding Paper Award out of 72 papers in international workshop.