Source author record

Jin Lu

Jin Lu appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Machine Learning Artificial Intelligence eess.SY Information Theory math.IT physics.app-ph physics.optics Systems and Control

Catalog footprint

What is connected

5works

8topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

DeepArrhythmia: Segment-Contextualized ECG Arrhythmia Classification via Selective Evidence Acquisition

Beat-level Electrocardiography (ECG) arrhythmia detection aims to assign an arrhythmia class to each beat in a recording, yet many existing systems treat beats as isolated local instances. This is limiting because beat labels often depend on multi-beat rhythm context, including timing, compensatory pauses, and beat-to-beat morphological consistency. We present DeepArrhythmia, a tool-grounded multimodal framework for segment-contextualized beat-level ECG arrhythmia classification. Given a multi-beat ECG segment, DeepArrhythmia combines the raw ECG signal and a rendered waveform image, localizes R peaks to identify beat instances, and produces structured beat-level predictions. The framework decouples physiological measurement from evidence integration using specialized tools for beat localization, numerical rhythm--morphology extraction, and morphology-focused textual analysis. DeepArrhythmia uses segment-level confidence to route between minimal and rich evidence states, since richer physiological evidence is not uniformly useful. This agentic design integrates rhythm context, explicit physiological grounding, and selective evidence acquisition for decision making.

preprint2022arXiv

The Benefits of Hydrogen Energy Transmission and Conversion Systems to the Renewable Power Grids: Day-ahead Unit Commitment

The curtailment of renewable energy is more frequently observed as the renewable penetration levels are rising rapidly in modern power systems. It is a waste of free and green renewable energy and implies current power grids are unable to accommodate more renewable sources. One major reason is that higher power transmission capacity is required for higher renewable penetration level. Another major reason is the volatility of the renewable generation. The hydrogen mix or pure hydrogen pipeline can both transfer and store the energy in the form of hydrogen. However, its potential of accelerating renewable integration has not been investigated. In this paper, hydrogen pipeline networks, combined with power-to-hydrogen (P2H) and hydrogen-to-power (H2P) facilities, are organized to form a Hydrogen Energy Transmission and Conversion System (HETCS). We investigate the operation of power systems coupled with HETCS, and propose the day-ahead security-constrained unit commitment (SCUC) with HETCS. The SCUC simulation is conducted on a modified IEEE 24-bus power system with HETCS. Simulation results show HETCS can substantially reduce the renewable curtailment, CO2 emission, load payment and total operational cost. This study validates the HETCS can be a promising solution to achieve net-zero renewable grids.

preprint2020arXiv

Experimental demonstration of multimode microresonator sensing by machine learning

A multimode microcavity sensor based on a self-interference microring resonator is demonstrated experimentally. The proposed multimode sensing method is implemented by recording wideband transmission spectra that consist of multiple resonant modes. It is different from the previous dissipative sensing scheme, which aims at measuring the transmission depth changes of a single resonant mode in a microcavity. Here, by combining the dissipative sensing mechanism and the machine learning algorithm, the multimode sensing information extracted from a broadband spectrum can be efficiently fused to estimate the target parameter. The multimode sensing method is immune to laser frequency noises and robust against system imperfection, thus our work presents a great step towards practical applications of microcavity sensors outside the research laboratory. The voltage applied across the microheater on the chip was adjusted to bring its influence on transmittance through the thermo-optic effects. As a proof-of-principle experiment, the voltage was detected by the multimode sensing approach. The experimental results demonstrate that the limit of detection of the multimode sensing by the general regression neural network is reduced to 6.7% of that of single-mode sensing within a large measuring range.

preprint2018arXiv

Edge Attention-based Multi-Relational Graph Convolutional Networks

Graph convolutional network (GCN) is generalization of convolutional neural network (CNN) to work with arbitrarily structured graphs. A binary adjacency matrix is commonly used in training a GCN. Recently, the attention mechanism allows the network to learn a dynamic and adaptive aggregation of the neighborhood. We propose a new GCN model on the graphs where edges are characterized in multiple views or precisely in terms of multiple relationships. For instance, in chemical graph theory, compound structures are often represented by the hydrogen-depleted molecular graph where nodes correspond to atoms and edges correspond to chemical bonds. Multiple attributes can be important to characterize chemical bonds, such as atom pair (the types of atoms that a bond connects), aromaticity, and whether a bond is in a ring. The different attributes lead to different graph representations for the same molecule. There is growing interests in both chemistry and machine learning fields to directly learn molecular properties of compounds from the molecular graph, instead of from fingerprints predefined by chemists. The proposed GCN model, which we call edge attention-based multi-relational GCN (EAGCN), jointly learns attention weights and node features in graph convolution. For each bond attribute, a real-valued attention matrix is used to replace the binary adjacency matrix. By designing a dictionary for the edge attention, and forming the attention matrix of each molecule by looking up the dictionary, the EAGCN exploits correspondence between bonds in different molecules. The prediction of compound properties is based on the aggregated node features, which is independent of the varying molecule (graph) size. We demonstrate the efficacy of the EAGCN on multiple chemical datasets: Tox21, HIV, Freesolv, and Lipophilicity, and interpret the resultant attention weights.

preprint2008arXiv

Linear Time Encoding of LDPC Codes

In this paper, we propose a linear complexity encoding method for arbitrary LDPC codes. We start from a simple graph-based encoding method ``label-and-decide.'' We prove that the ``label-and-decide'' method is applicable to Tanner graphs with a hierarchical structure--pseudo-trees-- and that the resulting encoding complexity is linear with the code block length. Next, we define a second type of Tanner graphs--the encoding stopping set. The encoding stopping set is encoded in linear complexity by a revised label-and-decide algorithm--the ``label-decide-recompute.'' Finally, we prove that any Tanner graph can be partitioned into encoding stopping sets and pseudo-trees. By encoding each encoding stopping set or pseudo-tree sequentially, we develop a linear complexity encoding method for general LDPC codes where the encoding complexity is proved to be less than $4 \cdot M \cdot (\overline{k} - 1)$, where $M$ is the number of independent rows in the parity check matrix and $\overline{k}$ represents the mean row weight of the parity check matrix.