Gram-negative bacteria are a major driver of antimicrobial resistance, yet most computational AMP methods are trained for broad-spectrum performance and do not capture outer-membrane specific constraints. We present SyAGram, a dual-component framework specialized for anti-Gram-negative peptide discovery. SyAGram-Synth is a discrete diffusion generator (D3PM) guided by a task-adapted ESM-2 encoder via crossattention, providing stable and controllable sequence synthesis. SyAGram-Pred is a hybrid classifier that fuses BiLSTM-based sequence features with global physicochemical descriptors to reflect charge and amphipathicity patterns critical for Gramnegative activity. On curated datasets, SyAGram-Pred attains 89.12% accuracy, 91.53% AUC, and 90.06% F1. SyAGram-Synth yields a mean predicted activity score of 0.856, with 92% of sequences predicted active ($\geq 0.5$) and 62.3 % at high confidence ($\geq 0.9$). Generated peptides exhibit physicochemical profiles consistent with natural AMPs. SyAGram thus offers a practical, targeted pipeline for peptide design against Gramnegative pathogens; a web demo is available at https://synergyg ram-4lsud.ondigitalocean.app/.

Transformer-based text-image matching model, known as CLIP, has garnered significant attention owing to its exceptional performance in text-image retrieval tasks and downstream applications. However, the interpret-ability of CLIP remains underexplored. Existing interpretation methods for Transformers often struggle with incomplete and unreliable attributions within the image and text modalities, respectively. In this paper, we propose a fine-grained interpretation method, termed Class Activation Values (CAV), to provide lucid and faithful visual explanations for CLIP-based text-image retrievals. Specifically, we systematically perform multi-scale accumulation and fusion of class-specific gradients and activation value features to generate high-definition explanations for the image encoder. Furthermore, we present element-wise gradient-based weights to attribute fine-grained relevance between value features and output similarity within the text encoder. The proposed CAV is capable of simultaneously rendering detailed and credible explanations due to its precise feature attribution. Extensive qualitative and quantitative experiments are conducted on the ImageNet-1k and MS COCO datasets, and the experimental results demonstrate that CAV outperforms state-of-the-art interpretation methods in both faithfulness and localization assessments across image and text modalities.

With advancements of diffusion models in generating high-quality, diverse, and creative content, subject-driven image editing based on diffusion models has been growing attention. However, it is challenging to perform precise control in image editing as minor changes in the input can lead to significant differences in generated concepts and content. Prevalent methods face significant limitations in terms of large numbers of trainable parameters and inference latency. Hence, we propose an efficient tensor-based fine-tuning method, Tucker/TT-based Fine-Tuning (TuckerFT and TTFT), for subject-driven image editing on diffusion models. Specifically, we first provide Tucker/TT-based Fine-tuning Convolutional Adapter and Attention Adapter to achieve parameter-efficient fine-tuning in convolutional and attention layers of diffusion models. Additionally, we further present a more sound initializing strategy for the Tucker/TT-based Adapters to guarantee steady training. Extensive experimental results demonstrate that TuckerFT/TTFT surpasses state-of-the-art methods on subject similarity, trainable parameters, and training duration.

As the visual interpretations for convolutional neural networks (CNNs), backpropagation attribution methods have been garnering growing attention. Nevertheless, majority of these methods merely concentrate on the ultimate convolutional layer, leading to tiny and concentrated interpretations that fail to adequately clarify the model-central attention. Therefore, we propose a precise attribution method (i.e., Holistic-CAM) for high-definition visual interpretation in the holistic stage of CNNs. Specifically, we first present weighted positive gradients to guarantee the sanity of interpretations in shallow layers and leverage multi-scale fusion to improve the resolution across the holistic stage. Then, we further propose a denoising strategy based on the fundamental scale component to eliminate the faithless attribution derived from fusing larger-scale features. The proposed method is capable of simultaneously rendering fine-grained and faithful interpretations for CNNs from shallow to deep layers. Extensive experimental results demonstrate that Holistic-CAM outperforms state-of-the-art methods on common-used benchmarks, including deletion and insertion, energy-based point game as well as Remove and Debias on ImageNet-1k, it also passes the sanity check easily.

In this paper, a rate-splitting multiple access (RSMA) scheme is proposed to aid a mobile edge computing (MEC) system where multiple randomly deployed users offload their computation tasks to a MEC server. Considering that users are divided into the center and edge groups, a cognitive radio (CR)-inspired rate-splitting is designed to enable the paired users to simultaneously offload to MEC server. Under the CR principles, the rate-splitting parameters are jointly designed to attain the maximum achievable rate for the secondary user, meanwhile maintaining the primary user’s offloading performance same as in orthogonal multiple access. For the case of the paired users with fixed locations, we derive a closed-form expression for the successful computation probability (SCP) achieved by the RSMA-aided MEC (RSMA-MEC) scheme and formulate a SCP maximization problem to obtain the optimal offloading parameters. To reveal the impact of the user locations on the offloading performance of the RSMA-MEC system, various distance-based user pairing schemes are investigated by invoking stochastic geometry techniques. We provide closed-form expressions for the SCPs achieved by the user pairing schemes to characterize the offloading performance. Pros and cons of the user locations to maximize the SCP are highlighted. Simulation results verify the accuracy of the analytical results and clarify the superior offloading performance achieved by the RSMA-MEC scheme, which attains a higher SCP than the existing schemes.

In this paper, we propose a composite multiple-mode orthogonal frequency division multiplexing with index modulation (C-MM-OFDM-IM) scheme to increase the spectral efficiency (SE) of OFDM-IM systems by extending the indexing to the energy and constellation domains. In C-MM-OFDM-IM, the information bits are mapped to not only the subcarrier activation patterns (SAPs) and modulation symbols, but also the energy allocation patterns (EAPs) and constellation activation patterns (CAPs). To cope with the practical situations, we propose a variant IM scheme named C-MM-OFDM-IM-II to build a new mapping rule between information bits and the increased CAPs, capable of further increasing the SE of C-MM-OFDM-IM. Upper-bounded bit error rate (BER) and lower-bounded achievable rate are both derived in closed-form to evaluate the performance of C-MM-OFDM-IM(-II). Moreover, we further propose two enhanced schemes, named generalized C-MM-OFDM-IM(-II) and C-MM-OFDM with in-phase/quadrature IM(-II), where the former jointly considers all SAPs, EAPs, CAPs and modulated symbols, while the latter expands the index implementation to the in-phase and quadrature constellation domains. Simulation results show that C-MM-OFDM-IM(-II) outperforms the conventional OFDM-IM related schemes, especially in the high signal-to-noise ratio (SNR) region, and verify the accuracy of the theoretical analysis for the upper-bounded BER and achievable rate.

Hyperspectral remote sensing image (HSI) consists of hundreds of bands that contain rich space, radiation and spectral information. The high-dimensional data can also lead to the curse of dimensionality problem making it difficult to be used effectively. In this paper, we proposed a manifold learning algorithm to reduce the dimensionality for HSI data. For high dimensional datasets with continuous variables, it is often the case that the data points are arranged along with low dimensional structures, named manifolds, in the high dimensional space. Manifold learning aims to identifying those special low dimensional structures for subsequent usage such as classification or regression. However, many manifold learning algorithms perform an eigenvector analysis on a data similarity matrix whose size is N×N, where N is the number of data points. The memory complexity of the analysis is at least O(N2) that is not feasible for a regular computer to compute or storage for very large datasets. To solve this problem, we used statistical sampling methods to sample a subset of data points as landmarks. A skeleton of the manifold was then identified based on the landmarks. The remaining data points were then inserted into the skeleton by Locally Linear Embedding (LLE). We tested our algorithm on AVIRIS Salinas-A data set. The experimental results showed that the HSI dataset could be reduced to a lower-dimensional space for land use classification with good performance, and the main structure was preserved well.