Recent advances in Neural Combinatorial Optimization (NCO) methods have significantly improved the capability of neural solvers to handle synthetic routing instances. Nonetheless, existing neural solvers typically struggle to generalize effectively from synthetic, uniformly-distributed training data to real-world VRP scenarios, including widely recognized benchmark instances from TSPLib and CVRPLib. To bridge this generalization gap, we present Evolutionary Realistic Instance Synthesis (EvoReal), which leverages an evolutionary module guided by large language models (LLMs) to generate synthetic instances characterized by diverse and realistic structural patterns. Specifically, the evolutionary module produces synthetic instances whose structural attributes statistically mimics those observed in authentic real-world instances. Subsequently, pre-trained NCO models are progressively refined, firstly aligning them with these structurally enriched synthetic distributions and then further adapting them through direct fine-tuning on actual benchmark instances. Extensive experimental evaluations demonstrate that EvoReal markedly improves the generalization capabilities of state-of-the-art neural solvers, yielding a notable reduced performance gap compared to the optimal solutions on the TSPLib (1.05%) and CVRPLib (2.71%) benchmarks across a broad spectrum of problem scales.

Deep learning has been extensively explored to solve vehicle routing problems (VRPs), which yields a range of data-driven neural solvers with promising outcomes. However, most neural solvers are trained to tackle VRP instances in a relatively monotonous context, e.g., simplifying VRPs by using Euclidean distance between nodes and adhering to a single problem size, which harms their off-the-shelf application in different scenarios. To enhance their versatility, this paper presents a novel lifelong learning framework that incrementally trains a neural solver to manage VRPs in distinct contexts. Specifically, we propose a lifelong learner (LL), exploiting a Transformer network as the backbone, to solve a series of VRPs. The inter-context self-attention mechanism is proposed within LL to transfer the knowledge obtained from solving preceding VRPs into the succeeding ones. On top of that, we develop a dynamic context scheduler (DCS), employing the cross-context experience replay to further facilitate LL looking back on the attained policies of solving preceding VRPs. Extensive results on synthetic and benchmark instances (problem sizes up to 18k) show that our LL is capable of discovering effective policies for tackling generic VRPs in varying contexts, which outperforms other neural solvers and achieves the best performance for most VRPs.

Acoustic resolution photoacoustic microscopy (AR-PAM) is a novel medical imaging modality, which can be used for both structural and functional imaging in deep bio-tissue. However, the imaging resolution is degraded and structural details are lost since its dependency on acoustic focusing, which significantly constrains its scope of applications in medical and clinical scenarios. To address the above issue, model-based approaches incorporating traditional analytical prior terms have been employed, making it challenging to capture finer details of anatomical bio-structures. In this paper, we proposed an innovative prior named group sparsity prior for simultaneous reconstruction, which utilizes the non-local structural similarity between patches extracted from internal AR-PAM images. The local image details and resolution are improved while artifacts are also introduced. To mitigate the artifacts introduced by patch-based reconstruction methods, we further integrate an external image dataset as an extra information provider and consolidate the group sparsity prior with a deep denoiser prior. In this way, complementary information can be exploited to improve reconstruction results. Extensive experiments are conducted to enhance the simulated and in vivo AR-PAM imaging results. Specifically, in the simulated images, the mean peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) values have increased from 16.36 dB and 0.46 to 27.62 dB and 0.92, respectively. The in vivo reconstructed results also demonstrate the proposed method achieves superior local and global perceptual qualities, the metrics of signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) have significantly increased from 10.59 and 8.61 to 30.83 and 27.54, respectively. Additionally, reconstruction fidelity is validated with the optical resolution photoacoustic microscopy (OR-PAM) data as reference image.

Deep reinforcement learning (DRL) frameworks have shown their remarkable effectiveness in learning navigation policy for the mobile robot navigating in a human crowded environment. Moreover, attention mechanisms coupled with DRL allows the robot to identify neighbors with different level of influence and incorporate them into the robot’s decision. However, as the crowd density increases, attention mechanisms may fail to identify critical neighbors which can lead to significant drops in navigation efficiency. In this work, we aim to address this limitation by encoding both human-human and human-robot interaction using a special class of Graph Convolutional Networks (GCN) known as Message-Passing GCN (MP-GCN). In contrast to existing methods, where attention between robot and humans are encoded uniformly, the proposed approach named MP-GatedGCN-RL encodes asymmetric interactions using the combination of novel message-passing function and edge-wise gating mechanisms. We evaluate our approach on the simulated environments of ETH/UCY pedestrians datasets consisting of different scenarios like collision avoidance, group forming, diverging, crossing, and so on. Experimental results demonstrate that our proposed method outperforms the conventional benchmark dynamic avoidance method ORCA with a 20.6% increase in success rate and a 9.1% reduction in navigation time. Moreover, we also achieve a 5.5% enhancement in success rate compared to other state-of-the-art DRL-based methods without any additional labeled expert data nor prior supervised learning.

Prediction of neutralization antibodies is important for the development of effective vaccines and antibody-based therapeutics. Traditional methods rely on features based on first principles derived from the binding interface. However, they are burdened by arduous data preprocessing from a limited quantity of protein structures. In comparison, deep learning allows automatic substructure characterization and representation without hand-crafted feature engineering. In particular, large language models (LLMs) based method predicts neutralization using Fv sequences of antibody and antigen. Despite LLM’s success, incorporating full-length Fv sequences suffers from: 1) inaccurate sequence-level labels in existing datasets, 2) inefficient modeling due to noisy non-contributing motifs, and 3) ignorance of non-bonded interactions that play a key role in facilitating epitope-paratope pairing. In this paper, we propose a novel approach that incorporates only the paratope and epitope for antibody-antigen neutralization prediction while adopting a novel set modeling that regards the paratope and epitope as bags of residues. Specifically, we hand-crafted a dataset containing neutralizing paratope-epitope pairs where epitopes are potentially generalizable to future unseen variants of SARS-CoV-2. Training on such a dataset enables deep learning models to predict neutralizing antibodies for prospective mutated variants of SARS-CoV-2, meanwhile addressing the problem of inaccurate sequence-level labels. A higher modeling efficiency is also achieved by disregarding non-contributing motifs. Furthermore, we also propose paratope-epitope set interaction (PESI), a set modeling model inspired by first principles that learns intra-inter non-covalent interactions through a global attention mechanism. To validate PESI, we perform a 10-fold cross-validation on our dataset. Experimental results show that PESI achieves a more balanced overall performance and a significant improvement on MCC as compared to existing architectures.

Representation learning-based recommendation models play a dominant role among recommendation techniques. However, most of the existing methods assume both historical interactions and embedding dimensions are independent of each other, and thus regrettably ignore the high-order interaction information among historical interactions and embedding dimensions. In this article, we propose a novel representation learning-based model called COMET (COnvolutional diMEnsion inTeraction), which simultaneously models the high-order interaction patterns among historical interactions and embedding dimensions. To be specific, COMET stacks the embeddings of historical interactions horizontally at first, which results in two “embedding maps”. In this way, internal interactions and dimensional interactions can be exploited by convolutional neural networks (CNN) with kernels of different sizes simultaneously. A fully connected multi-layer perceptron (MLP) is then applied to obtain two interaction vectors. Lastly, the representations of users and items are enriched by the learnt interaction vectors, which can further be used to produce the final prediction. Extensive experiments and ablation studies on various public implicit feedback datasets clearly demonstrate the effectiveness and rationality of our proposed method.

Re-ranking models refine item recommendation lists generated by the prior global ranking model, which have demonstrated their effectiveness in improving the recommendation quality. However, most existing re-ranking solutions only learn from implicit feedback with a shared prediction model, which regrettably ignore inter-item relationships under diverse user intentions. In this paper, we propose a novel Intention-aware Re-ranking Model with Dynamic TransformerEncoder (RAISE), aiming to perform user-specific prediction for each individual user based on her intentions. Specifically, we first propose to mine latent user intentions from text reviews with an intention discovering module (IDM). By differentiating the importance of review information with a co-attention network, the latent user intention can be explicitly modeled for each user-item pair. We then introduce a dynamic transformer encoder (DTE) to capture user-specific inter-item relationships among item candidates by seamlessly accommodating the learned latent user intentions via IDM. As such, one can not only achieve more personalized recommendations but also obtain corresponding explanations by constructing RAISE upon existing recommendation engines. Empirical study on four public datasets shows the superiority of our proposed RAISE, with up to 13.95%, 9.60%, and 13.03% relative improvements evaluated by Precision@5, MAP@5, and NDCG@5 respectively.

The rapid evolution of influenza viruses constantly leads to the emergence of novel influenza strains that are capable of escaping from population immunity. The timely determination of antigenic variants is critical to vaccine design. Empirical experimental methods like hemagglutination inhibition (HI) assays are time-consuming and labor-intensive, requiring live viruses. Recently, many computational models have been developed to predict the antigenic variants without considerations of explicitly modeling the interdependencies between the channels of feature maps. Moreover, the influenza sequences consisting of similar distribution of residues will have high degrees of similarity and will affect the prediction outcome. Consequently, it is challenging but vital to determine the importance of different residue sites and enhance the predictive performance of influenza antigenicity. We have proposed a 2D convolutional neural network (CNN) model to infer influenza antigenic variants (IAV-CNN). Specifically, we apply a new distributed representation of amino acids, named ProtVec that can be applied to a variety of downstream proteomic machine learning tasks. After splittings and embeddings of influenza strains, a 2D squeeze-and-excitation CNN architecture is constructed that enables networks to focus on informative residue features by fusing both spatial and channel-wise information with local receptive fields at each layer. Experimental results on three influenza datasets show IAV-CNN achieves state-of-the-art performance combining the new distributed representation with our proposed architecture. It outperforms both traditional machine algorithms with the same feature representations and the majority of existing models in the independent test data. Therefore we believe that our model can be served as a reliable and robust tool for the prediction of antigenic variants.

MOTIVATION: Influenza viruses are persistently threatening public health, causing annual epidemics and sporadic pandemics. The evolution of influenza viruses remains to be the main obstacle in the effectiveness of antiviral treatments due to rapid mutations. Previous work has been investigated to reveal the determinants of virulence of the influenza A virus. To further facilitate flu surveillance, explicit detection of influenza virulence is crucial to protect public health from potential future pandemics. RESULTS: In this article, we propose a weighted ensemble convolutional neural network (CNN) for the virulence prediction of influenza A viruses named VirPreNet that uses all eight segments. Firstly, mouse lethal dose 50 is exerted to label the virulence of infections into two classes, namely avirulent and virulent. A numerical representation of amino acids named ProtVec is applied to the eight-segments in a distributed manner to encode the biological sequences. After splittings and embeddings of influenza strains, the ensemble CNN is constructed as the base model on the influenza dataset of each segment, which serves as the VirPreNet's main part. Followed by a linear layer, the initial predictive outcomes are integrated and assigned with different weights for the final prediction. The experimental results on the collected influenza dataset indicate that VirPreNet achieves state-of-the-art performance combining ProtVec with our proposed architecture. It outperforms baseline methods on the independent testing data. Moreover, our proposed model reveals the importance of PB2 and HA segments on the virulence prediction. We believe that our model may provide new insights into the investigation of influenza virulence. AVAILABILITY AND IMPLEMENTATION: Codes and data to generate the VirPreNet are publicly available at https://github.com/Rayin-saber/VirPreNet. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Positive-unlabeled (PU) learning handles the problem of learning a predictive model from PU data. Past few years have witnessed the boom of PU learning, while the existing learning algorithms are limited to binary classification and cannot be directly applied to multi-class PU data. In this paper, we present an unbiased estimator of the original classification risk for multi-class PU learning, and show that the direct empirical risk minimization suffers from the severe overfitting problem because the risk is unbounded below. To address this problem, we propose an alternative risk estimator, and theoretically establish an estimation error bound. We show that the estimation error of its empirical risk minimizer achieves the optimal parametric convergence rate. Extensive experiments on multiple datasets demonstrate the effectiveness of the proposed approach for multi-class PU learning.

Trained with the standard cross entropy loss, deep neural networks can achieve great performance on correctly labeled data. However, if the training data is corrupted with label noise, deep models tend to overfit the noisy labels, thereby achieving poor generation performance. To remedy this issue, several loss functions have been proposed and demonstrated to be robust to label noise. Although most of the robust loss functions stem from Categorical Cross Entropy (CCE) loss, they fail to embody the intrinsic relationships between CCE and other loss functions. In this paper, we propose a general framework dubbed Taylor cross entropy loss to train deep models in the presence of label noise. Specifically, our framework enables to weight the extent of fitting the training labels by controlling the order of Taylor Series for CCE, hence it can be robust to label noise. In addition, our framework clearly reveals the intrinsic relationships between CCE and other loss functions, such as Mean Absolute Error (MAE) and Mean Squared Error (MSE). Moreover, we present a detailed theoretical analysis to certify the robustness of this framework. Extensive experimental results on benchmark datasets demonstrate that our proposed approach significantly outperforms the state-of-the-art counterparts.

In the era of big data, traditional supply chain systems can not match the requirement of e-commerce. The analysis of customers’ demands and behaviors are necessary to exploit the potential insights and to build intelligent supply chain systems, which can be achieved by recommender systems. Graph-based recommendation models work well for top-N recommender systems due to their capability to capture the potential relationships between entities. In this paper, we propose a novel graph-based recommendation model to achieve personalized item ranking. To be specific, we design an adapted semi-supervised learning method to capture item smoothness, item fitting, and item confidence. By exploiting the structure of item graph moderately, the proposed method achieves impressive effectiveness and efficiency. In addition, extensive experimental results on real-world datasets show that our proposed method consistently outperforms the state-of-the-art counterparts on the top-N recommendation task.