Source author record

Teng Long

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

ResearcherUnclaimed source record

Computation and Language cond-mat.mes-hall physics.optics

Catalog footprint

What is connected

2works

3topics

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.

preprint2024arXiv

Optical spin Hall effect pattern switching in polariton condensates in organic single-crystal microbelts

Topological polaritons, combining the robustness of the topological protected edge states to defects and disorder with the strong nonlinear properties of polariton bosons, represent an excellent platform to investigate novel photonic topological phases. In this work, we demonstrated the optical spin Hall effect (OSHE) and its symmetry switching in the exciton-polariton regime of pure DPAVBi crystals. Benefiting from the photonic Rashba-Dresselhaus spin-orbit coupling in organic crystals, we observed the separation of left- and right-circularly-polarized polariton emission in two-dimensional momentum space and real space, a signature of the OSHE. Above the lasing threshold, the OSHE pattern changes due to transverse quantization in the microbelt. This device without superlattice structure has great potential applications in topological polaritonics, such as information transmission, photonic integrated chips and quantum information.

preprint2016arXiv

Leveraging Lexical Resources for Learning Entity Embeddings in Multi-Relational Data

Recent work in learning vector-space embeddings for multi-relational data has focused on combining relational information derived from knowledge bases with distributional information derived from large text corpora. We propose a simple approach that leverages the descriptions of entities or phrases available in lexical resources, in conjunction with distributional semantics, in order to derive a better initialization for training relational models. Applying this initialization to the TransE model results in significant new state-of-the-art performances on the WordNet dataset, decreasing the mean rank from the previous best of 212 to 51. It also results in faster convergence of the entity representations. We find that there is a trade-off between improving the mean rank and the hits@10 with this approach. This illustrates that much remains to be understood regarding performance improvements in relational models.