Search

Pretraining molecular representations is crucial for drug and material discovery. Recent methods focus on learning representations from geometric structures, effectively capturing 3D position information. Yet, they overlook the rich information in biomedical texts, which detail molecules' properties and substructures. With this in mind, we set up a data collection effort for 200K pairs of ground-state geometric structures and biomedical texts, resulting in a PubChem3D dataset. Based on this dataset, we propose the GeomCLIP framework to enhance for multi-modal representation learning from molecular structures and biomedical text. During pre-training, we design two types of tasks, i.e., multimodal representation alignment and unimodal denoising pretraining, to align the 3D geometric encoder with textual information and, at the same time, preserve its original representation power. Experimental results show the effectiveness of GeomCLIP in various tasks such as molecular property prediction, zero-shot text-molecule retrieval, and 3D molecule captioning. Our code and collected dataset are available at \url{https://github.com/xiaocui3737/GeomCLIP}, Comment: BIBM 2024

This paper introduces a novel generalized self-imitation learning ($\textbf{GSIL}$) framework, which effectively and efficiently aligns large language models with offline demonstration data. We develop $\textbf{GSIL}$ by deriving a surrogate objective of imitation learning with density ratio estimates, facilitating the use of self-generated data and optimizing the imitation learning objective with simple classification losses. $\textbf{GSIL}$ eliminates the need for complex adversarial training in standard imitation learning, achieving lightweight and efficient fine-tuning for large language models. In addition, $\textbf{GSIL}$ encompasses a family of offline losses parameterized by a general class of convex functions for density ratio estimation and enables a unified view for alignment with demonstration data. Extensive experiments show that $\textbf{GSIL}$ consistently and significantly outperforms baselines in many challenging benchmarks, such as coding (HuamnEval), mathematical reasoning (GSM8K) and instruction-following benchmark (MT-Bench)., Comment: EMNLP 2024 Main

Recent advancements in biology and chemistry have leveraged multi-modal learning, integrating molecules and their natural language descriptions to enhance drug discovery. However, current pre-training frameworks are limited to two modalities, and designing a unified network to process different modalities (e.g., natural language, 2D molecular graphs, 3D molecular conformations, and 3D proteins) remains challenging due to inherent gaps among them. In this work, we propose MolBind, a framework that trains encoders for multiple modalities through contrastive learning, mapping all modalities to a shared feature space for multi-modal semantic alignment. To facilitate effective pre-training of MolBind on multiple modalities, we also build and collect a high-quality dataset with four modalities, MolBind-M4, including graph-language, conformation-language, graph-conformation, and conformation-protein paired data. MolBind shows superior zero-shot learning performance across a wide range of tasks, demonstrating its strong capability of capturing the underlying semantics of multiple modalities.

This paper proposes a low-cost six Degree-of-Freedom (6-DOF) navigation system for small aerial robots based on the integration of Global Position System (GPS) receiver with sensors of inertional Microelectromechanical Systems (MEMS). In the problem of fusing Inertial Measurement Unit (IMU) with low-cost GPS, the effect of time synchronization error on attitude estimation is concerned. A fusion algorithm which can estimate the motion states and the time synchronization error simultaneously is proposed. This algorithm adds a time estimation loop to improve estimation accuracy. Compared with another states augmented estimation approach, this method has the advantages of lower computation burden, avoidance of the discretization error in the low sample rate. The estimation algorithm is implemented in an low-cost embedded microprocessor where the update rate of algorithm can achieve more than 100 Hz, and therefore high-performance computational units are not necessary. In robotic experiment, the proposed algorithm serves as the navigation solution for a small aerial robot. The accuracy and reliability of the self-designed system are tested when the robot is moving with significant acceleration., Comment: 9 pages, 12 Figures

With the help of quantum key distribution (QKD), two distant peers are able to share information-theoretically secure key bits. Increasing key rate is ultimately significant for the applications of QKD in lossy channel. However, it has proved that there is a fundamental rate-distance limit, named linear bound, which limits the performance of all existing repeaterless protocols and realizations. Surprisingly, a recently proposed protocol, called twin-field (TF) QKD can beat linear bound with no need of quantum repeaters. Here, we present the first implementation of TF-QKD protocol and demonstrate its advantage of beating linear bound at the channel distance of 300 km. In our experiment, a modified TF-QKD protocol which does not assume phase post-selection is considered, and thus higher key rate than the original one is expected. After well controlling the phase evolution of the twin fields travelling hundreds of kilometers of optical fibres, the implemented system achieves high-visibility single-photon interference, and allows stable and high-rate measurement-device-independent QKD. Our experimental demonstration and results confirm the feasibility of the TF-QKD protocol and its prominent superiority in long distance key distribution services.

Among the various versions of the twin-field quantum key distribution (TF-QKD) protocol [M.Lucamarini, Z. Yuan, J. Dynes, and A. Shields, Nature (London) 557, 400 (2018)] that can overcome the rate-distance limit, the TF-QKD without phase postselection proposed by Cui et al. [Phys. Rev. Appl. 11, 034053 (2019)] is an elegant TF-QKD that can provide high key rates since the postselection of global phases has been removed. However, the achievable distance of this variant is shorter than that of the original phase-matching QKD [X. Ma, P. Zeng, and H. Zhou, Phys. Rev. X 8, 031043 (2018)]. In this paper, we propose a method for improving its performance by introducing an additional decoy mode. The upper bound of the information leakage can be more tightly estimated; hence, both the key rate and the achievable distance are significantly improved. Interestingly, the operation of the proposed additional decoy mode is the same as that of the code mode; hence, it does not introduce difficulties into the experimental system. In addition, the improvement is substantial with finite decoy states, which is meaningful in practice., Comment: 6 pages, 3 figures

The parameters choosing (such as probabilities of choosing X-basis or Z-basis, intensity of signal state and decoy state, etc.) and system calibrating will be more challenging when the number of users of a measurement-device-independent quantum key distribution(MDI-QKD) network becomes larger. At present, people usually use optimization algorithms to search the best parameters. This method can find the optimized parameters accurately but may cost lots of time and hardware resources. It's a big problem in large scale MDI-QKD network. Here, we present a new method, using Back Propagation Artificial Neural Network(BPNN) to predict, rather than searching the optimized parameters. Compared with optimization algorithms, our BPNN is faster and more lightweight, it can save system resources. Another big problem brought by large scale MDI-QKD network is system recalibration. BPNN can support this work in real time, and it only needs to use some discarded data generated from communication process, rather than require us to add additional devices or scan the system, Comment: 7 pages, 8 figures