arXiv subjects

Qiming Chen

Publications and source records attributed to Qiming Chen.

7 recordsLinked to original sources

arXiv2026

Baguan-TS: A Sequence-Native In-Context Learning Model for Time Series Forecasting with Covariates

Transformers enable in-context learning (ICL) for rapid, gradient-free adaptation in time series forecasting, yet most ICL-style approaches rely on tabularized, hand-crafted features, while end-to-end sequence models lack inference-time adaptation. We bridge this gap with a unified framework, Baguan-TS, which integrates the raw-sequence representation learning with ICL, instantiated by a 3D Transformer that attends jointly over temporal, variable, and context axes. To make this high-capacity model practical, we tackle two key hurdles: (i) calibration and training stability, improved with a feature-agnostic, target-space retrieval-based local calibration; and (ii) output oversmoothing, mitigated via context-overfitting strategy. On public benchmark with covariates, Baguan-TS consistently outperforms established baselines, achieving the highest win rate and significant reductions in both point and probabilistic forecasting metrics. Further evaluations across diverse real-world energy datasets demonstrate its robustness, yielding substantial improvements.

cs.LG↗

arXiv2025

TriForecaster: A Mixture of Experts Framework for Multi-Region Electric Load Forecasting with Tri-dimensional Specialization

Electric load forecasting is pivotal for power system operation, planning and decision-making. The rise of smart grids and meters has provided more detailed and high-quality load data at multiple levels of granularity, from home to bus and cities. Motivated by similar patterns of loads across different cities in a province in eastern China, in this paper we focus on the Multi-Region Electric Load Forecasting (MRELF) problem, targeting accurate short-term load forecasting for multiple sub-regions within a large region. We identify three challenges for MRELF, including regional variation, contextual variation, and temporal variation. To address them, we propose TriForecaster, a new framework leveraging the Mixture of Experts (MoE) approach within a Multi-Task Learning (MTL) paradigm to overcome these challenges. TriForecaster features RegionMixer and Context-Time Specializer (CTSpecializer) layers, enabling dynamic cooperation and specialization of expert models across regional, contextual, and temporal dimensions. Based on evaluation on four real-world MRELF datasets with varied granularity, TriForecaster outperforms state-of-the-art models by achieving an average forecast error reduction of 22.4\%, thereby demonstrating its flexibility and broad applicability. In particular, the deployment of TriForecaster on the eForecaster platform in eastern China exemplifies its practical utility, effectively providing city-level, short-term load forecasts for 17 cities, supporting a population exceeding 110 million and daily electricity usage over 100 gigawatt-hours.

cs.LG↗

arXiv2024

Decay-protected superconducting qubit with fast control enabled by integrated on-chip filters

Achieving fast gates and long coherence times for superconducting qubits presents challenges, typically requiring either a stronger coupling of the drive line or an excessively strong microwave signal to the qubit. To address this, we introduce on-chip filters of the qubit drive exhibiting a stopband at the qubit frequency, thus enabling long coherence times and strong coupling at the subharmonic frequency, facilitating fast single-qubit gates, and reduced thermal load. The filters exhibit an extrinsic relaxation time of a few seconds while enabling sub-10-ns gates with subharmonic control. Here we show up to 200-fold improvement in the measured relaxation time at the stopband. Furthermore, we implement subharmonic driving of Rabi oscillations with a $\pi$ pulse duration of 12 ns. Our demonstration of on-chip filters and efficient subharmonic driving in a two-dimensional quantum processor paves the way for a scalable qubit architecture with reduced thermal load and noise from the control line.

quant-ph↗

arXiv2023

Single-Shot Readout of a Superconducting Qubit Using a Thermal Detector

Measuring the state of qubits is one of the fundamental operations of a quantum computer. Currently, state-of-the-art high-fidelity single-shot readout of superconducting qubits relies on parametric amplifiers at the millikelvin stage. However, parametric amplifiers are challenging to scale beyond hundreds of qubits owing to practical size and power limitations. Nanobolometers have a multitude of properties that are advantageous for scalability and have recently shown sensitivity and speed promising for qubit readout, but such thermal detectors have not been demonstrated for this purpose. In this work, we utilize an ultrasensitive bolometer in place of a parametric amplifier to experimentally demonstrate single-shot qubit readout. With a readout duration of $13.9~μ\mathrm{s}$, we achieve a single-shot fidelity of 0.618 which is mainly limited by the energy relaxation time of the qubit, $T_1 = 28~μ\mathrm{s}$. Without the $T_1$ errors, we find the fidelity to be 0.927. In the future, high-fidelity single-shot readout may be achieved by straightforward improvements to the chip design and experimental setup, and perhaps most interestingly by the change of the bolometer absorber material to reduce the readout time to the hundred-nanosecond level and beyond.

quant-ph↗

arXiv2022

S3E: A Multi-Robot Multimodal Dataset for Collaborative SLAM

The burgeoning demand for collaborative robotic systems to execute complex tasks collectively has intensified the research community's focus on advancing simultaneous localization and mapping (SLAM) in a cooperative context. Despite this interest, the scalability and diversity of existing datasets for collaborative trajectories remain limited, especially in scenarios with constrained perspectives where the generalization capabilities of Collaborative SLAM (C-SLAM) are critical for the feasibility of multi-agent missions. Addressing this gap, we introduce S3E, an expansive multimodal dataset. Captured by a fleet of unmanned ground vehicles traversing four distinct collaborative trajectory paradigms, S3E encompasses 13 outdoor and 5 indoor sequences. These sequences feature meticulously synchronized and spatially calibrated data streams, including 360-degree LiDAR point cloud, high-resolution stereo imagery, high-frequency inertial measurement units (IMU), and Ultra-wideband (UWB) relative observations. Our dataset not only surpasses previous efforts in scale, scene diversity, and data intricacy but also provides a thorough analysis and benchmarks for both collaborative and individual SLAM methodologies. For access to the dataset and the latest information, please visit our repository at https://pengyu-team.github.io/S3E.

cs.RO↗

arXiv2022

Perspectives of microwave quantum key distribution in open-air

One of the cornerstones of quantum communication is the unconditionally secure distribution of classical keys between remote parties. This key feature of quantum technology is based on the quantum properties of propagating electromagnetic waves, such as entanglement, or the no-cloning theorem. However, these quantum resources are known to be susceptible to noise and losses, which are omnipresent in open-air communication scenarios. In this work, we theoretically investigate the perspectives of continuous-variable open-air quantum key distribution at microwave frequencies. In particular, we present a model describing the coupling of propagating microwaves with a noisy environment. Using a protocol based on displaced squeezed states, we demonstrate that continuous-variable quantum key distribution with propagating microwaves can be unconditionally secure at room temperature up to distances of around 200 meters. Moreover, we show that microwaves can potentially outperform conventional quantum key distribution at telecom wavelength at imperfect weather conditions.

quant-ph↗

arXiv2017

CNN Is All You Need

The Convolution Neural Network (CNN) has demonstrated the unique advantage in audio, image and text learning; recently it has also challenged Recurrent Neural Networks (RNNs) with long short-term memory cells (LSTM) in sequence-to-sequence learning, since the computations involved in CNN are easily parallelizable whereas those involved in RNN are mostly sequential, leading to a performance bottleneck. However, unlike RNN, the native CNN lacks the history sensitivity required for sequence transformation; therefore enhancing the sequential order awareness, or position-sensitivity, becomes the key to make CNN the general deep learning model. In this work we introduce an extended CNN model with strengthen position-sensitivity, called PoseNet. A notable feature of PoseNet is the asymmetric treatment of position information in the encoder and the decoder. Experiments shows that PoseNet allows us to improve the accuracy of CNN based sequence-to-sequence learning significantly, achieving around 33-36 BLEU scores on the WMT 2014 English-to-German translation task, and around 44-46 BLEU scores on the English-to-French translation task.

cs.CL↗