This study introduces a data-driven surrogate modelling framework that combines an artificial neural network (ANN) with particle swarm optimisation (PSO) and a genetic algorithm (GA) to optimise methanol production under uncertain conditions. A steady-state Aspen Plus model was developed and converted into dynamic mode by applying ± 5% uncertainty across 12 key process variables, generating 3880 data points that reflect realistic operational variability. The ANN model was trained and validated on the samples, achieving predictive accuracy (R ² = 0.988, RMSE = 28.59) on unseen test data. Key features of the work include the use of the ANN as a surrogate model, its integration within PSO and GA optimisation frameworks and its application alongside Sobol and Fourier amplitude sensitivity test (FAST) methods to identify the most influential process variables affecting the methanol production rate. The proposed framework resulted in performance improvements, with PSO achieving an increase of 38.63% and GA 33.14% in methanol production. Cross-validation with the Aspen Plus model confirmed the reliability of the optimised operating conditions, with relative errors ranging from 0.07% to 2.15%. Overall, the study demonstrates the effectiveness of integrating surrogate modelling with intelligent optimisation techniques to improve the efficiency and robustness of methanol production processes under uncertainty.

Security operations centre (SOC) analysts must investigate alerts, correlate threat intelligence and interpret heterogeneous telemetry under tight timing constraints. Although large language models (LLMs) offer strong understanding capabilities, directly applying them to SOC environments remains challenging due to semantic ambiguity in analyst queries, fragmented multisource event data, limited domain-specific reasoning and reliability concerns associated with unconstrained query generation. We present a task-driven knowledge-augmented framework designed to produce verifiable and contextually grounded responses for SOC workflows. The framework integrates four components: (i) contrastive context task recognition that mitigates semantic ambiguity by mapping analyst queries to predefined SOC task types; (ii) expert-guided knowledge augmentation that fuses dense and sparse retrieval to bridge the semantic gap; (iii) schema-aligned event retrieval combined with entity-centric evidence profiling to ensure reliable and secure access to heterogeneous telemetry and (iv) verifiable task-aware generation that constrains model outputs to retrieved knowledge and security events. To assess the framework, we construct a benchmark of 12,500 validated question–answer pairs derived through semiautomated synthesis over more than 34 million real SOC records. Experiments across multiple foundation models demonstrate consistent improvements in relevance and grounding quality. Our results indicate that the four proposed components substantially enhance LLMs' reliability in practical SOC analysis.

Leveraging both global contextual dependencies and local temporal-spectral patterns can further enhance speech quality and intelligibility, motivating the integration of diverse neural network structures for improved mask estimation. Furthermore, due to the limitations of existing time-frequency phase-aware masks, a new constrained phase-sensitive mask is introduced and estimated using the proposed architectures. In this paper, we first propose a modified transformer model for constrained phase-sensitive ideal magnitude ratio mask (cPSIRM) estimation (TRF-cPSIRM) to incorporate both magnitude and phase information and improve speech enhancement quality. The transformer model can extract global information, whereas the convolutional neural network (CNN) and convolutional recurrent network (CRN) are effective in capturing local content. CNNs have feature extraction capability due to their convolutional layers, and CRNs benefit from the temporal modelling strength of LSTM recurrent layers, which is useful for enhancement. Therefore, in this paper, to exploit both transformer (TRF) and CNN/CRN capabilities, we propose the cascaded structures (models) of CNN and TRF layers (CNN-TRF-cPSIRM) and also TRF and CRN layers (TRF-CRN-cPSIRM) for cPSIRM estimation and, consequently, speech enhancement. The CNN is used for feature extraction in CNN-TRF-cPSIRM and CRN for enhancement in TRF-CRN-cPSIRM. Moreover, considering the harmonic property of speech in the frequency spectrum, we present the transposed transformer-based model (TTRF), in which the neighbourhood relationship between the different frequency sub-bands is used as a sequence in the model. Then, to model both the long-term and short-term dependencies, the cascaded model of TTRF (intra-transformer) and TRF (inter-transformer) is proposed for cPSIRM estimation (TTRF-TRF-cPSIRM). The experimental results show that among these transformer-based models, the CNN-TRF-cPSIRM has the best performance, achieving up to about 0.44 perceptual evaluation of speech quality (PESQ) improvement over the baseline and 1.5 over the noisy speech.

Parameter-efficient fine-tuning (PEFT) has become a crucial paradigm for domain adaptation, achieving strong performance by updating only a small fraction of model parameters. Among various PEFT methods, low-rank adaptation (LoRA) is widely adopted due to its structural simplicity and computational efficiency. However, in multitask scenarios, LoRA often suffers from performance degradation due to task interference. Recent extensions, such as mixture-of-experts (MoE) and asymmetric LoRA variants, attempt to mitigate these issues; however, their reliance on fixed subspace-mixing strategies limits flexibility and makes the models more sensitive to input noise and data sparsity. Moreover, vanilla LoRA typically initialises low-rank matrices with Gaussian noise or zeros and optimises them in unconstrained subspaces, which may disrupt the structured representations learnt by pretrained models. In this article, we propose M3LoRA, a novel PEFT framework that leverages multiple low-rank matrices with mixture-of-subspaces and minor singular components initialisation. M3LoRA utilises multiple low-rank matrices to minimise interference between task-specific subspaces while maintaining representational capacity. Furthermore, it employs a learnable mixing matrix positioned between the down-projection and up-projection matrices to dynamically combine their subspaces, thereby enabling adaptive task-specific combination mechanisms. Additionally, it initialises these low-rank matrices within a subspace orthogonal to the principal singular components of pretrained weights-termed the minor singular components-thereby leveraging directions unexplored during pretraining to better capture task-specific features from labelled data. Extensive experiments on a wide range of benchmark datasets demonstrate that M3LoRA achieves substantial improvements over existing PEFT baselines, particularly in multi-task scenarios where task interference often degrades conventional LoRA performance.

Building energy systems integrating multiple energy sources can effectively reduce energy consumption and facilitate renewable energy integration. Integrating electrical energy storage (EES) into these systems helps accommodate the increasing share of renewables; however, the stochastic and intermittent nature of solar power still poses challenges to supply reliability. This study proposes a photovoltaic (PV)-oriented storage scheduling strategy, in which short-term PV generation forecasts are applied to guide the operation of a building power supply network consisting of photovoltaic panels, the grid, and energy storage systems. The forecasting approach employs a hybrid framework combining a Long Short-Term Memory (LSTM) network to capture temporal dependencies, an attention mechanism to emphasise critical time steps, and a Temporal Convolutional Network (TCN) to map the enhanced features to PV outputs. Experimental evaluation using historical datasets under multiple weather conditions and time periods shows that the proposed LSTM-Attention-TCN model achieves a mean absolute error (MAE) of 20.45 W/m² and a Nash–Sutcliffe efficiency (NSE) of 0.94, outperforming both standalone LSTM and TCN models as well as their hybrid variants in terms of accuracy and robustness. By providing high-accuracy solar irradiance forecasts to guide energy storage operation and grid interaction, the proposed model enables more efficient and economical scheduling of building energy systems. Compared with an uncontrolled scenario, the LSTM-Attention-TCN-based scheduling reduces the total operating cost by approximately 52.1%, and achieves an additional 16.5% reduction compared to a conventional strategy without predictive coordination. In addition, compared to other hybrid forecasting models such as LSTM-TCN and TCN-Attention, the proposed model achieves the lowest total cost of CNY 14.83 and demonstrates superior scheduling efficiency, thereby enhancing the stability and flexibility of building energy utilization.

Underwater image enhancement remains a critical challenge in computational vision due to complex distortions caused by wavelength-dependent light absorption and scattering. This paper introduces CEDFNet, a novel two-stage framework that leverages advanced computational intelligence techniques for robust and high-fidelity underwater image restoration. The first stage integrates a Colour Equalisation Transformer (CET) to perform global colour correction by modelling long-range dependencies and mitigating dominant hue distortions. The second stage combines a Residual Texture Modulation Adaptor (RTMA) with an Enhanced Bilateral Enhancement Decoder (EBED) to refine structural details and enhance local contrast through context-aware and adaptive feature learning. Extensive evaluations on benchmark datasets including UIEBD, LSUI, and Colour-Checker7 validate the superiority of CEDFNet over existing state-of-the-art approaches. Quantitatively, CEDFNet achieves significant improvements across multiple perceptual and fidelity metrics such as PSNR, SSIM, FID, and LPIPS. Comprehensive ablation studies further confirm the complementary roles of CET, RTMA, and EBED, whereas parameter sensitivity analyses highlight the framework's robust and stable behaviour. By integrating transformer-based global correction with task-adaptive local enhancement, CEDFNet advances the frontier of underwater image restoration in the domain of computational intelligence. It generalises well across diverse imaging conditions and offers a lightweight and end-to-end solution suitable for real-world deployment in marine robotics, inspection, and visual perception systems.

Extracting spatio-temporal cues from neighbouring frames is challenging in video super-resolution (VSR). Although deformable alignment-based VSR methods have shown promise in aligning neighbouring frames with the reference frame, most existing methods rely on one or a few traditional convolutions to estimate motion offsets for spatio-temporal alignment, restricting receptive field size and alignment accuracy. To address these limitations, we propose an effective spatio-temporal alignment network (ESTA-Net) for VSR. The core component of our method is the group convolution-based alignment module (GCBAM), which utilises cascaded group convolutions to learn offsets across both the original and downsampled resolutions. By employing group convolutions rather than traditional convolutions, GCBAM enables the deformable alignment to achieve a wider receptive field with lower computational cost, thereby improving the accuracy of offset estimation. Additionally, the bi-scale alignment strategy within GCBAM enhances robustness to complex and large-scale motions. Furthermore, we introduce an attention-based feature enhancement module (AFEM) to refine the aligned features, focusing on critical details to improve reconstruction quality. Extensive experiments on standard benchmarks show that our ESTA-Net achieves superior VSR performance against other advanced methods, while maintaining a good equilibrium between model size and performance.

Large language models (LLMs) have made remarkable advances in natural language processing, demonstrating great potential in modelling structured sequences. However, adapting these capabilities to machine gaming tasks such as Go remains challenging due to limitations in strategy generalisation and optimisation efficiency. This paper presents multitype game optimisation (MyGO), a two-stage fine-tuning framework tailored for two-player perfect information board games, exploring the applicability of LLMs to nonlinguistic decision-making domains. In the supervised fine-tuning stage, we propose a unified structural encoding method, action semantic unit (ASU), which efficiently converts heterogeneous game records into discrete token sequences compatible with LLMs. In the reinforcement learning stage, we design TA-PPO (token-level adaptive proximal policy optimisation), an enhanced PPO-based algorithm to address the issue of sparse feedback commonly encountered in game reinforcement learning. Experimental results demonstrate that the fine-tuned models achieve superior or comparable performance to traditional game-playing algorithms in terms of strategy quality, rule generalisation and inference efficiency. This work provides a scalable paradigm for fine-tuning LLMs in complex decision-making tasks and lays a foundation for future research in game AI and generalisable strategy optimisation.

Partial Multi-label Learning (PML) deals with the ambiguity where each instance is annotated with a set of candidate labels, and only a subset of which is valid. While existing PML methods focus primarily on label disambiguation, they often rely on the assumption of a clean feature space. However, in real-world applications, data are frequently plagued by the co-existence of label noise and feature noise, referred to as the dual noise challenge. Consequently, model robustness degrades substantially. To address this, we propose a framework named Ranking-Consistent Correntropy-based subspace learning for Partial Multi-label Learning (RCC-PML). Unlike existing dual noise PML methods that operate in the input space, our work introduces a subspace learning framework, where robust representation and semantic ranking are jointly optimized to enforce cross-space consistency. Specifically, we leverage the Maximum Correntropy Criterion (MCC) to construct robust scatter matrices, effectively suppressing heavy-tailed feature noise. To tackle label ambiguity, a ranking-consistent constraint is introduced to encourage a reasonable margin between ground-truth and false-positive labels in the projected subspace. Furthermore, we incorporate dual-graph regularization to preserve both the local manifold structure via anchor embedding and global semantic consistency. Finally,L_2,1-norm regularization is imposed on the projection matrix to perform adaptive feature selection. Extensive experiments on benchmark datasets demonstrate that the proposed method significantly outperforms state-of-the-art algorithms, particularly in heavy-tailed environments.

Adversarial training (AT) is widely regarded as a crucial defense method for deep neural networks against adversarial attacks. Most of the existing AT methods suffer from the problems of insufficient coverage of perturbation space and robust overfitting. In view of this, we propose an AT framework with adaptive example reuse (AT-AER) to help improve the adversarial robustness of deep models. In AT-AER, a new concept named 2nd-order adversarial example (AE) is proposed by adaptively filtering AEs generated during the historical training phase, which achieves sufficient coverage of diverse attack directions. Meanwhile, by analysing the fundamental causes of robust overfitting, we propose the strategies of wave descending learning rate (WDLR), cosine increasing weight decay (CIWD) and cosine increasing attack strength (CIAS) in collaboration with AT-AER to optimise models. In addition, the Stochastic Weight Averaging (SWA) technique is introduced to further improve the stability of training. Finally, experiments on three benchmark datasets show that AT-AER exhibits significant advantages in the face of strong adversarial attacks. Its adaptive mechanism effectively alleviates the phenomenon of robust overfitting where the performance difference between the best model and the last model is less than 1%. The study further reveals that using traditional weak attacks (e.g., FGSM) to evaluate the robustness of models may lead to a false sense of reliability, indicating the necessity of using strong attacks for robustness evaluation. This study provides a solution for AT that balances efficiency and performance.

In this paper, to deal with the dynamic SLAM problem, we investigate feature tracking and IMU preintegration in visual-inertial odometry (VIO) and design a robust SLAM framework that explicitly considers robot self-dynamics. We propose a self-dynamics and IMU-aided feature tracker to predict initial optical flow and an iterative refinement method that accounts for patch affine deformation and illumination changes, improving tracking accuracy and robustness. Furthermore, we introduce an SE₂(3)-based IMU preintegration that preserves state correlations and consistently encodes robot self-dynamics for subsequent optimisation. A VIO framework with preprocessing, optimisation and loop-closing threads is developed to validate the proposed self-dynamics–aware tracker and SE₂(3)-based preintegration. Experiments, including module tests and ablation studies, demonstrate improved feature tracking accuracy, IMU noise propagation and overall VIO performance when explicitly modelling robot self-dynamics.

Models based on U-shaped networks have achieved widespread success in the field of medical image segmentation, but their performance is generally limited by structural bottlenecks in the network. At this stage, feature maps experience a sharp decline in spatial resolution due to continuous downsampling, resulting in significant loss of critical boundaries and structural details. Additionally, the local receptive fields of convolutions limit the effective modelling of global context. To address this core issue, we propose a novel enhanced segmentation network called FMTNet. FMTNet fundamentally enhances the expressive power of deep features by integrating an innovative composite enhancement module at the bottleneck of the U-Net. This module consists of three synergistically working submodules: the Fourier spatial fusion module, which introduces a frequency-domain perspective to compensate for and reconstruct high-frequency structural information lost in the spatial domain; the hybrid mamba–transformer module, which efficiently captures cross-regional long-range dependencies to establish global context and the multi-scale context Aggregation module, which fuses features of different scales to adapt to objects of varying sizes. We conducted extensive experiments on multiple public multi-modal datasets, including colonoscopy polyps, dermatoscopy lesions, breast ultrasound and dental X-rays. The results demonstrate that FMTNet comprehensively outperforms SOTA methods across all key metrics, showcasing exceptional segmentation accuracy and generalisation capabilities. Our research study demonstrates that by synergistically enhancing deep features across three dimensions-frequency, global, and multi-scale-FMTNet provides a general and efficient solution to address the bottleneck issues of U-Net, significantly enhancing the accuracy and robustness of medical image segmentation. The source code and pre-trained weights are available at https://github.com/shiguiling0-has/FMTNet.

The deployment of deep neural networks (DNNs) in safety-critical domains is critically hampered by their vulnerability to defects, which can arise from malicious attacks or low-quality data. Therefore, precisely locating the network components responsible for these defects, and subsequently repairing them without compromising overall model performance, presents a significant challenge. To address this, this paper introduces NSRepair, a framework that combines interpretable fault localisation with multi-objective optimisation. Specifically, to accurately attribute blame for a defect, we employ Shapley values to quantify the contribution of each neuron. To systematically manage the trade-off between defect correction and performance preservation, we formulate the repair task as a multi-objective optimisation problem. We conducted extensive experiments across four distinct repair tasks, validating NSRepair on diverse model architectures against seven specialised state-of-the-art methods. The results demonstrate that our unified framework effectively repairs a wide range of defects, demonstrating its potential as a versatile and practical solution for improving DNN dependability. Our code is publicly available at https://doi.org/10.5281/zenodo.17494304.

Standard deep learning optimisation is typically conducted on shape-fixed loss surfaces. However, shape-fixed loss surfaces may impede optimisers from reaching flat regions closely associated with strong generalisation. In this work, we propose a new paradigm named deformation mapping to deform the loss surface during optimisation. Moreover, we design various vertical deformation mappings (VDMs) and further analyse their contributions to the training process. Theoretically, we prove that deforming the loss surface enhances the optimiser's ability to filter out sharp minima in deterministic settings. Furthermore, by incorporating diffusion theory, we demonstrate that VDM exponentially reduces the escape time from sharp minima under stochastic noise and momentum. Empirically, visualisations of loss landscapes demonstrate that VDMs locate significantly flatter minima compared to standard optimisation. Furthermore, integrating VDMs into the training of various deep neural networks produces consistent accuracy gains on ImageNet, CIFAR-10, and CIFAR-100, with negligible additional computation. Notably, PreResNet-20 on CIFAR-100 achieves a 1.46% increase in top-1 accuracy. These results indicate that the deformation mapping is a promising paradigm for improving optimisation and generalisation in deep learning. The code is available at https://anonymous.4open.science/r/Vertical-Deformation-Mapping-2324.

Papillary Thyroid Carcinoma (PTC) is the most prevalent thyroid malignancy, and accurate lesion segmentation is essential for clinical diagnosis and treatment planning. Metaheuristic optimisation algorithms have been widely used in Multi-Threshold Image Segmentation (MTIS), but many existing methods suffer from an imbalance between global exploration and local exploitation. This study aims to develop a robust and well-balanced optimisation algorithm to improve the accuracy and stability of MTIS for PTC images. An Adaptive Guided Polar Lights Optimisation (AGPLO) algorithm is proposed, which incorporates an adaptive phase-shift operator, magnetic guiding convergence, and energy burst exploration mechanisms to dynamically regulate search behaviour. AGPLO was evaluated on the IEEE CEC2017 benchmark suite and applied to Rényi entropy-based MTIS for PTC image segmentation. Experimental results on benchmark functions demonstrate that AGPLO outperforms several original and advanced metaheuristic algorithms in terms of convergence accuracy, stability, and robustness. In PTC image segmentation experiments, AGPLO achieves superior PSNR, SSIM, and FSIM values, producing clearer lesion boundaries and preserving structural details more effectively than comparative methods. The proposed AGPLO provides an effective and reliable optimisation framework for MTIS and shows strong potential for intelligent medical image analysis applications.

Traffic Flow Forecasting (TFF) is a foundational task in the development of Intelligent Transport Systems (ITSs). The primary challenge is to undertake a comprehensive exploration of the intrinsic dynamic spatiotemporal correlations of the road network, unveiling the long-term evolutionary traffic trends. Furthermore, most existing methods often solely depend on the single traffic condition and neglect the enhancement of correlated features collected from traffic sensors in prediction. To this end, we propose a dynamic correlation-information-fusion-based (DCIF) spatiotemporal network for TFF, which models the spatiotemporal correlations of road networks, thereby effectively capturing dynamically changing characteristics. Specifically, a spatiotemporal feature enhancement (STFE) mechanism is employed to capture the directional and location-aware characteristics of traffic flow, thereby enhancing the representation of traffic flow and the capability of spatiotemporal feature extraction. Then, a gated attention unit (GAU) is constructed to meticulously extract the deep dynamic trends inherent within traffic data. Finally, a dynamic feature matrix (DFM) is formulated, incorporating spatial graph convolution to provide comprehensive semantic contextual information. The DFM captures the dynamic topology of the deeper feature network in real time by fusing spatial node information and traffic speed features as correlation information. Extensive experiments demonstrate that DCIF significantly outperforms other baselines in prediction accuracy, thereby further substantiating its validity and reliability in TFF.

Open-vocabulary 3D querying based on 3D Gaussian splatting (3DGS) shows great promise in facilitating accurate 3D query capabilities of AI systems. These methods typically rely on pre-captured multi-view images to enable natural language interactions with 3D scenes. In practice, when embodied AI encounters unexplored scenes, it is difficult to obtain observations from different viewpoints beforehand. This challenge highlights the importance of exploring natural language-driven 3D scene querying from a single current viewpoint. This paper proposes single view language Gaussian splatting (SVLGaussian) for the novel task: Open-vocabulary 3D querying based on the input single view. By leveraging multi-round inference of multimodal large language models, SVLGaussian efficiently generates pixel-level semantic probabilities and rapidly embeds them into a 3D Gaussian field, enabling real-time language-guided semantic querying. To verify our model, we annotated three datasets: Lerf_ovs and 3D-OVS, which are tailored for open-vocabulary 3D querying, and RE10K, which is adapted for single-view 3D reconstruction. Both quantitative and qualitative results show that our method effectively supports open-vocabulary 3D querying from a single view.

Accurate segmentation of colorectal polyps is essential for early colorectal cancer screening, yet remains challenging due to weak foreground–background contrast, disrupted boundaries caused by specular reflections and intestinal folds, and pronounced scale variation among polyps. These factors make it difficult for existing methods to jointly preserve fine boundary details and robust global semantic context. To address these task-specific challenges, we propose a Dual-branch Feature Progressive Fusion Network (DFPF-Net) for colorectal polyp segmentation. DFPF-Net adopts a dual-encoder architecture that integrates a CNN-based encoder for local and boundary-sensitive representation for global semantic modelling. A boundary-aware branch equipped with stacked Inversely Perceive Information Layers (IPILs) enhances ambiguous and fragmented contours, while the semantic branch incorporates Misalignment Fusion Modules (MFMs) and a Misaligned Single-layer Reinforcement Module (MSRM) to alleviate semantic misalignment and insufficient cross-scale interaction. Furthermore, a Perceptual Information Fusion Module (PIFM) enables effective semantic–boundary collaboration, and a Multi-level Residual Decoding Module (MRDM) progressively reconstructs structurally consistent segmentation outputs. Extensive experiments on multiple public colonoscopy datasets demonstrate that DFPF-Net achieves competitive and robust segmentation performance. In particular, on the challenging ETIS dataset, DFPF-Net attains 0.785 mDice and 0.704 mIoU, indicating its capability in handling complex structures and ambiguous boundaries in colorectal polyp segmentation.

Networked control systems (NCSs) often suffer from performance degradation due to limited communication bandwidth, which can cause data transmission conflicts and packet loss. Existing scheduling strategies may fail to simultaneously meet the real-time requirements and the importance of multisensor data, and they are particularly vulnerable under distributed denial of service (DDoS) attacks. Firstly, to address these challenges, a greedy algorithm is proposed to optimise the data transmission process to satisfy the importance and real-time requirement of sensor data. It enables the dynamic allocation of network resources, reduces the possibility of data conflict and improves the communication efficiency. Then, an observer-based control algorithm is designed to ensure the system's stability and security resilience under bandwidth constraints and DDoS attacks. Therefore, in the face of packet loss caused by data conflict, the control algorithm can maintain efficient resource scheduling capability and improve the robustness of the system. Finally, the simulation results show that the proposed dynamic resource scheduling framework can guarantee the performance of NCSs under constrained network conditions.

In recent years, prompt learning has shown promise in transferring pretrained vision-language models (VLMs) to downstream tasks. However, existing methods face two challenges in improving generalisation: (1) When leveraging the collaborative effect of multimodal prompts, it is often assumed that text and visual modalities share the same prompt requirements, neglecting the distinct hierarchical processing of their encoders, leading to prompt imbalance; and (2) current methods exhibit limited adaptability when facing diverse distribution shift scenarios, including class distribution shifts and image content variations. To address these challenges, we propose a diversified composite prompt learning (DCPL) framework that integrates unified and specific prompts. Specifically, to alleviate multimodal prompt imbalance, we design a shared root multimodal prompting strategy, which employs a shared root prompt and an independent derivation mechanism to generate the derived multimodal prompt (DMP), enabling independent deep prompting while maintaining implicit synergy across modalities. Furthermore, we design a dual-branch dynamic adaptive prompting strategy that produces the derived class-specific prompt (DCP) and image-specific prompt (ISP), driven by inter-class relations and image-patch context, respectively, to enhance adaptability across different distribution shifts. Extensive experiments on base-to-novel, cross-dataset, domain generalisation and few-shot learning demonstrate that the DCPL achieves superior performance, validating its robustness and generalisation.