Shi Ying

Research profile

Research summary

Shi Ying (Wuhan University, China; ORCID 0000-0002-0471-0021) is an early-career researcher working at the intersection of cloud computing systems and machine-learning-based scheduling. Because 'Shi Ying' is a common name, this profile is name-resolved from public bibliographic data linked to that ORCID. The record shows a steep activity ramp: 1 indexed work in each of 2023 and 2024, then 7 in 2025 and 7 in 2026 — 15 works over the last three years. Topic coverage is led by Software System Performance and Reliability (11), alongside Anomaly Detection (6), Software Engineering Research (5), Network Security and Intrusion Detection (5), and Cloud Computing and Resource Management (2). Evolve has reviewed two 2026 arXiv preprints, both with Shi Ying in a middle, non-corresponding position: HiGFRL, a hierarchical graph fusion reinforcement-learning scheduler for dependency-aware DAG tasks in heterogeneous clouds, and MCRL2, a cross-attention-augmented soft actor-critic method for microservice scheduling. Both carry medium reproducibility readiness (record mean 55), with Evolve flagging simulation-only evaluation, absent variance testing, and single-run results. Citation impact is not yet measurable (h-index 2, i10-index 0), and the collaboration footprint is a compact 4-person coauthor network. Overall: a fast-rising, narrowly focused scheduling researcher whose method designs look promising but whose experimental validation is not yet at publication-grade rigor.

Research keywords

Highlights

• Hierarchical graph fusion DRL for DAG scheduling in heterogeneous clouds

  The 2026 HiGFRL preprint combines hypergraph and multi-view graph fusion with an actor-critic (PPO) reinforcement learner for dependency-aware task scheduling on heterogeneous cloud resources. Evolve judges the design and ablation studies solid, but notes the evaluation is simulation-only and reports no variance testing.

• Cross-attention augmented SAC for cloud microservice scheduling

  The 2026 MCRL2 preprint adds multi-resource cross-attention representation learning to soft actor-critic for cloud microservice scheduling and load balancing. Evolve finds it beats DRL baselines in trace-driven simulation, while cautioning that results come from a single run and the cross-node claim is shaky.

• A consistent evaluation-rigor gap across both reviewed papers

  Both reviewed 2026 papers are rated medium for reproducibility (record mean 55), and both verdicts converge on the same weakness: single-run, simulation-only experiments without variance testing. Addressing statistical rigor and releasing code or traces would be the clearest upgrade path for this line of work.

Timeline

1. 2023First indexed work appears in the OpenAlex-linked record, with 1 work in this year.
2. 2024Output remains minimal at 1 indexed work, before the current productive phase begins.
3. 2025Indexed output rises sharply to 7 works, marking the start of a sustained productive phase.
4. 2026A further 7 works, including the two Evolve-reviewed arXiv preprints HiGFRL and MCRL2 on reinforcement-learning-based cloud scheduling.

Papers in the Evolve catalog

• 2026-09-14

  HiGFRL: Hierarchical Graph Fusion-Driven Reinforcement Learning for Dependency-Aware Task Scheduling in Heterogeneous Cloud

• 2026-09-11

  MCRL2: Multi-resource Cross-attention-based Representation Learning-augmented Reinforcement Learning for Cloud Microservice Scheduling

Frequent co-authors