Tengfei Liu

Tengfei Liu

Auto-generated profile, based on publicly available information

Tengfei Liu

Department of Computer Science and Engineering, Southern University of Science and Technology, Shenzhen, China

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
20 September 2024 in IEEE Transactions on Evolutionary Computation
Verified
Journal Article
Closed access
QZQi Zhao
Qi Zhao
TLTengfei Liu
Tengfei Liu
BYBai Yan
Bai Yan
QDQiqi Duan
Qiqi Duan
JYJian Yang
Jian Yang
YSYuhui Shi
Yuhui Shi
Automated design of metaheuristic algorithms offers an attractive avenue to reduce human effort and gain enhanced performance beyond human intuition. Current automated methods design algorithms within a fixed structure and operate from scratch. This poses a clear gap towards fully discovering potentials over the metaheuristic family and fertilizing from prior design experience. To bridge the gap, this paper proposes an autoregressive learning-based designer for automated design of metaheuristic algorithms. Our designer formulates metaheuristic algorithm design as a sequence generation task, and harnesses an autoregressive generative network to handle the task. This offers two advances. First, through autoregressive inference, the designer generates algorithms with diverse lengths and structures, enabling to fully discover potentials over the metaheuristic family. Second, prior design knowledge learned and accumulated in neurons of the designer can be retrieved for designing algorithms for future problems, paving the way to continual design of algorithms for open-ended problem-solving. Extensive experiments on numeral benchmarks and real-world problems reveal that the proposed designer generates algorithms that outperform all human-created baselines on 24 out of 25 test problems. The generated algorithms display various structures and behaviors, reasonably fitting for different problem-solving contexts. Code is available at https://github.com/auto4opt/ALDeshttps://github.com/auto4opt/ALDes.

23 June 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
TLTengfei Liu
Tengfei Liu
JLJinrun Liu
Jinrun Liu
BYBai Yan
Bai Yan
QZQi Zhao
Qi Zhao
YSYuhui Shi
Yuhui Shi
Node selection critically affects the efficiency of branch-and-bound (BnB) for mixed-integer linear programming (MILP). Existing learning-based methods usually score active nodes independently, which ignores the hierarchical search structure and often leads to policies that rely on scale-dependent cues. We propose TreeCL, a tree-aware contrastive learning framework that reformulates node selection as learning structural relations between candidate nodes and their ancestral context. TreeCL is based on an intrinsic property of BnB, termed optimal-path consistency: if a globally optimal solution remains feasible at a descendant node, it must also remain feasible at all its ancestors. This property enables principled supervision from BnB trajectories by contrasting on-path and off-path nodes with respect to their lowest common ancestor. The resulting similarity function supports efficient node comparison and can be integrated into standard BnB execution. We further introduce a branching-sequence-based augmentation strategy to improve sample efficiency under limited expert supervision. Experiments on multiple MILP benchmarks show that TreeCL reduces explored nodes and improves solving time compared with heuristic and learning-based baselines. It also transfers more robustly to larger instances and improves solution quality on real-world datasets.