Zhen Li

Bio: Zhen Li is an academic researcher from Wuhan University. The author has contributed to research in topics: Medicine & Computer science. The author has an hindex of 127, co-authored 1712 publications receiving 71351 citations. Previous affiliations of Zhen Li include Tsinghua University & Hong Kong University of Science and Technology.

Knowledge Concentration: Learning 100K Object Classifiers in a Single CNN.

Abstract: Fine-grained image labels are desirable for many computer vision applications, such as visual search or mobile AI assistant. These applications rely on image classification models that can produce hundreds of thousands (e.g. 100K) of diversified fine-grained image labels on input images. However, training a network at this vocabulary scale is challenging, and suffers from intolerable large model size and slow training speed, which leads to unsatisfying classification performance. A straightforward solution would be training separate expert networks (specialists), with each specialist focusing on learning one specific vertical (e.g. cars, birds...). However, deploying dozens of expert networks in a practical system would significantly increase system complexity and inference latency, and consumes large amounts of computational resources. To address these challenges, we propose a Knowledge Concentration method, which effectively transfers the knowledge from dozens of specialists (multiple teacher networks) into one single model (one student network) to classify 100K object categories. There are three salient aspects in our method: (1) a multi-teacher single-student knowledge distillation framework; (2) a self-paced learning mechanism to allow the student to learn from different teachers at various paces; (3) structurally connected layers to expand the student network capacity with limited extra parameters. We validate our method on OpenImage and a newly collected dataset, Entity-Foto-Tree (EFT), with 100K categories, and show that the proposed model performs significantly better than the baseline generalist model.

Copolymer of Fluorene and Triphenylamine Moieties: Direct and Post‐Functionalization Strategy, Structural Characterization, and Chemosensoring Behavior

Abstract: To prepare a copolymer of fluorene and triphenylamine bearing imidazole groups in the side chain, two synthetic approaches have been designed: the direct copolymerization of the corresponding monomers, and the post-functionalization of a precursor polymer. Unexpectedly, the direct copolymeriztion strategy was better. The chemosensing behavior of P2 was carefully studied, the fluorescence of which could be quenched by Cu 2+ ions at concentrations as low as 8 x 10 -6 mol· L -1 in dilute solutions. By utilizing the much higher stability constant of the complex of CN-and Cu 2+ , the quenched fluorescence of the solution of P2 by Cu 2+ ions could be recovered upon the addition of a trace amount of CN - anions, with a detection limit as low as 1.8 x 10 -5 mol · L -1 (0.47 ppm), making P2 a new sensitive cyanide probe.

SOI high-voltage device with step thickness sustained voltage layer

Abstract: A new SOI high-voltage device with a step thickness sustained voltage layer (ST SOI) is proposed. The electric field in the drift region is modulated, and that in the buried layer is enhanced by the variable-thickness SOI layer, resulting in enhancement of breakdown voltage (BV). BV for the ST SOI with two steps is twice as high as that of the conventional SOI, maintaining the low on-resistance (Ron).

Facile one-step vulcanization of copper foil towards stable Li metal anode

Abstract: Lithium metal has been regarded as the ultimate anode for next-generation rechargeable batteries with high energy density. However, its high reactivity and dendrite growth seriously limit its commercial application, which can be well addressed by realizing uniform Li deposition. Here, we report a facile and scalable one-step vulcanization method to modify commercial Cu foil with lithophilic Cu2S. The in situ formed Cu2S layer can not only promote the homogeneous deposition of Li via its lithophilic nature, but also benefit the formation of a stable solid-electrolyte interphase during initial activation. The Cu2S-modified Cu current collector realizes dendrite-free Li plating/stripping and thus exhibits stable cycling performance with a high Coulombic efficiency, even with a large capacity of 4 mA h cm−2. A full-cell consisting of a Cu2S/Cu-Li anode and a LiFePO4 cathode exhibits greatly improved cycling stability and enhanced Coulombic efficiency, demonstrating the effectiveness and practicability of the proposed Cu2S/Cu foil in the field of rechargeable Li metal batteries.

Preparation of Co–N–C supported on silica spheres with high catalytic performance for ethylbenzene oxidation

Abstract: In this study, Co–N–C supported on silica spheres is prepared through heat-treatment of supported metalloporphyrin in an N2 atmosphere. The catalytic performance of the catalysts for ethylbenzene oxidation is investigated and these catalysts are characterized by techniques such as BET, FT-IR, UV-vis, TEM and XPS. In comparison with other catalysts such as supported cobalt porphyrin and unsupported cobalt porphyrin, Co–N–C supported on silica spheres shows much higher catalytic activity for ethylbenzene oxidation (15.7%) and selectivity to acetophenone (76.5%). In addition, the catalyst can retain its high catalytic activity after several reuses. The results show that the high catalytic performance of the catalyst may be attributed to the formation of Co–N4–C sites during the heat-treatment and supported Co–N–C catalysts may be beneficial to yield more Co–N4–C sites.

I and i

Abstract: There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality. Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

Fast parallel algorithms for short-range molecular dynamics

Abstract: Three parallel algorithms for classical molecular dynamics are presented. The first assigns each processor a fixed subset of atoms; the second assigns each a fixed subset of inter-atomic forces to compute; the third assigns each a fixed spatial region. The algorithms are suitable for molecular dynamics models which can be difficult to parallelize efficiently—those with short-range forces where the neighbors of each atom change rapidly. They can be implemented on any distributed-memory parallel machine which allows for message-passing of data between independently executing processors. The algorithms are tested on a standard Lennard-Jones benchmark problem for system sizes ranging from 500 to 100,000,000 atoms on several parallel supercomputers--the nCUBE 2, Intel iPSC/860 and Paragon, and Cray T3D. Comparing the results to the fastest reported vectorized Cray Y-MP and C90 algorithm shows that the current generation of parallel machines is competitive with conventional vector supercomputers even for small problems. For large problems, the spatial algorithm achieves parallel efficiencies of 90% and a 1840-node Intel Paragon performs up to 165 faster than a single Cray C9O processor. Trade-offs between the three algorithms and guidelines for adapting them to more complex molecular dynamics simulations are also discussed.

Highly accurate protein structure prediction with AlphaFold

Abstract: Proteins are essential to life, and understanding their structure can facilitate a mechanistic understanding of their function. Through an enormous experimental effort1–4, the structures of around 100,000 unique proteins have been determined5, but this represents a small fraction of the billions of known protein sequences6,7. Structural coverage is bottlenecked by the months to years of painstaking effort required to determine a single protein structure. Accurate computational approaches are needed to address this gap and to enable large-scale structural bioinformatics. Predicting the three-dimensional structure that a protein will adopt based solely on its amino acid sequence—the structure prediction component of the ‘protein folding problem’8—has been an important open research problem for more than 50 years9. Despite recent progress10–14, existing methods fall far short of atomic accuracy, especially when no homologous structure is available. Here we provide the first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known. We validated an entirely redesigned version of our neural network-based model, AlphaFold, in the challenging 14th Critical Assessment of protein Structure Prediction (CASP14)15, demonstrating accuracy competitive with experimental structures in a majority of cases and greatly outperforming other methods. Underpinning the latest version of AlphaFold is a novel machine learning approach that incorporates physical and biological knowledge about protein structure, leveraging multi-sequence alignments, into the design of the deep learning algorithm. AlphaFold predicts protein structures with an accuracy competitive with experimental structures in the majority of cases using a novel deep learning architecture.