Bingyu Liu

Youjun Xu, Jinchuan Xiao, Chia-Han Chou et al.

Molecular structures are always depicted as 2D printed form in scientific documents like journal papers and patents. However, these 2D depictions are not machine-readable. Due to a backlog of decades and an increasing amount of these printed literature, there is a high demand for the translation of printed depictions into machine-readable formats, which is known as Optical Chemical Structure Recognition (OCSR). Most OCSR systems developed over the last three decades follow a rule-based approach where the key step of vectorization of the depiction is based on the interpretation of vectors and nodes as bonds and atoms. Here, we present a practical software MolMiner, which is primarily built up using deep neural networks originally developed for semantic segmentation and object detection to recognize atom and bond elements from documents. These recognized elements can be easily connected as a molecular graph with distance-based construction algorithm. We carefully evaluate our software on four benchmark datasets with the state-of-the-art performance. Various real application scenarios are also tested, yielding satisfactory outcomes. The free download links of Mac and Windows versions are available: Mac: https://molminer-cdn.iipharma.cn/pharma-mind/artifact/latest/mac/PharmaMind-mac-latest-setup.dmg and Windows: https://molminer-cdn.iipharma.cn/pharma-mind/artifact/latest/win/PharmaMind-win-latest-setup.exe

Yuan Ge, Haishu Zhao, Aokai Hao et al.

Emotion is a core paralinguistic feature in voice interaction. It is widely believed that emotion understanding models learn fundamental representations that transfer to synthesized speech, making emotion understanding results a plausible reward or evaluation metric for assessing emotional expressiveness in speech synthesis. In this work, we critically examine this assumption by systematically evaluating Speech Emotion Recognition (SER) on synthesized speech across datasets, discriminative and generative SER models, and diverse synthesis models. We find that current SER models can not generalize to synthesized speech, largely because speech token prediction during synthesis induces a representation mismatch between synthesized and human speech. Moreover, generative Speech Language Models (SLMs) tend to infer emotion from textual semantics while ignoring paralinguistic cues. Overall, our findings suggest that existing SER models often exploit non-robust shortcuts rather than capturing fundamental features, and paralinguistic understanding in SLMs remains challenging.

Chunyu Sun, Bingyu Liu, Zhichao Cui et al.

Embedding-based retrieval models have made significant strides in retrieval-augmented generation (RAG) techniques for text and multimodal large language models (LLMs) applications. However, when it comes to speech larage language models (SLLMs), these methods are limited to a two-stage process, where automatic speech recognition (ASR) is combined with text-based retrieval. This sequential architecture suffers from high latency and error propagation. To address these limitations, we propose a unified embedding framework that eliminates the need for intermediate text representations. Specifically, the framework includes separate speech and text encoders, followed by a shared scaling layer that maps both modalities into a common embedding space. Our model reduces pipeline latency by 50\% while achieving higher retrieval accuracy compared to traditional two-stage methods. We also provide a theoretical analysis of the challenges inherent in end-to-end speech retrieval and introduce architectural principles for effective speech-to-document matching. Extensive experiments demonstrate the robustness of our approach across diverse acoustic conditions and speaker variations, paving the way for a new paradigm in multimodal SLLMs retrieval systems.

Li Guo, Yonghong Peng, Rui Qin et al.

Iron ore feed load control is one of the most critical settings in a mineral grinding process, directly impacting the quality of final products. The setting of the feed load is mainly determined by the characteristics of the ore pellets. However, the characterisation of ore is challenging to acquire in many production environments, leading to poor feed load settings and inefficient production processes. This paper presents our work using deep learning models for direct ore feed load estimation from ore pellet images. To address the challenges caused by the large size of a full ore pellets image and the shortage of accurately annotated data, we treat the whole modelling process as a weakly supervised learning problem. A two-stage model training algorithm and two neural network architectures are proposed. The experiment results show competitive model performance, and the trained models can be used for real-time feed load estimation for grind process optimisation.

Shangyu Xie, Bingyu Liu, Yuan Hong

Deep learning as a service (DLaaS) has been intensively studied to facilitate the wider deployment of the emerging deep learning applications. However, DLaaS may compromise the privacy of both clients and cloud servers. Although some privacy preserving deep neural network (DNN) based inference techniques have been proposed by composing cryptographic primitives, the challenges on computational efficiency have not been well-addressed due to the complexity of DNN models and expensive cryptographic primitives. In this paper, we propose a novel privacy preserving cloud-based DNN inference framework (namely, "PROUD"), which greatly improves the computational efficiency. Finally, we conduct extensive experiments on two commonly-used datasets to validate both effectiveness and efficiency for the PROUD, which also outperforms the state-of-the-art techniques.

Bingyu Liu, Yuhong Guo, Jieping Ye et al.

Recent domain adaptation methods have demonstrated impressive improvement on unsupervised domain adaptation problems. However, in the semi-supervised domain adaptation (SSDA) setting where the target domain has a few labeled instances available, these methods can fail to improve performance. Inspired by the effectiveness of pseudo-labels in domain adaptation, we propose a reinforcement learning based selective pseudo-labeling method for semi-supervised domain adaptation. It is difficult for conventional pseudo-labeling methods to balance the correctness and representativeness of pseudo-labeled data. To address this limitation, we develop a deep Q-learning model to select both accurate and representative pseudo-labeled instances. Moreover, motivated by large margin loss's capacity on learning discriminative features with little data, we further propose a novel target margin loss for our base model training to improve its discriminability. Our proposed method is evaluated on several benchmark datasets for SSDA, and demonstrates superior performance to all the comparison methods.

Zhen Zhao, Bingyu Liu, Yuhong Guo et al.

In this paper, we present our proposed ensemble model with batch spectral regularization and data blending mechanisms for the Track 2 problem of the cross-domain few-shot learning (CD-FSL) challenge. We build a multi-branch ensemble framework by using diverse feature transformation matrices, while deploying batch spectral feature regularization on each branch to improve the model's transferability. Moreover, we propose a data blending method to exploit the unlabeled data and augment the sparse support set in the target domain. Our proposed model demonstrates effective performance on the CD-FSL benchmark tasks.

Bingyu Liu, Zhen Zhao, Zhenpeng Li et al.

In this paper, we propose a feature transformation ensemble model with batch spectral regularization for the Cross-domain few-shot learning (CD-FSL) challenge. Specifically, we proposes to construct an ensemble prediction model by performing diverse feature transformations after a feature extraction network. On each branch prediction network of the model we use a batch spectral regularization term to suppress the singular values of the feature matrix during pre-training to improve the generalization ability of the model. The proposed model can then be fine tuned in the target domain to address few-shot classification. We also further apply label propagation, entropy minimization and data augmentation to mitigate the shortage of labeled data in target domains. Experiments are conducted on a number of CD-FSL benchmark tasks with four target domains and the results demonstrate the superiority of our proposed model.

Yuan Hong, Han Wang, Shangyu Xie et al.

Energy has been increasingly generated or collected by different entities on the power grid (e.g., universities, hospitals and householdes) via solar panels, wind turbines or local generators in the past decade. With local energy, such electricity consumers can be considered as "microgrids" which can simulataneously generate and consume energy. Some microgrids may have excessive energy that can be shared to other power consumers on the grid. To this end, all the entities have to share their local private information (e.g., their local demand, local supply and power quality data) to each other or a third-party to find and implement the optimal energy sharing solution. However, such process is constrained by privacy concerns raised by the microgrids. In this paper, we propose a privacy preserving scheme for all the microgrids which can securely implement their energy sharing against both semi-honest and colluding adversaries. The proposed approach includes two secure communication protocols that can ensure quantified privacy leakage and handle collusions.