Intelligent technologies are driving the development of smart campuses, fostering a dynamic and diverse intelligent ***, the current trend of customizing smart campus solutions often positions campus citizens as mere ...
详细信息
Deep Neural Networks (DNNs) have found successful applications in various non-safety-critical domains. However, given the inherent lack of interpretability in DNNs, ensuring their prediction accuracy through robustnes...
详细信息
ISBN:
(纸本)9783031664557;9783031664564
Deep Neural Networks (DNNs) have found successful applications in various non-safety-critical domains. However, given the inherent lack of interpretability in DNNs, ensuring their prediction accuracy through robustness verification becomes imperative before deploying them in safety-critical applications. Neural Network Verification (NNV) approaches can broadly be categorized into exact and approximate solutions. Exact solutions are complete but time-consuming, making them unsuitable for large network architectures. In contrast, approximate solutions, aided by abstraction techniques, can handle larger networks, although they may be incomplete. This paper introduces AccMILP, an approach that leverages abstraction to transform NNV problems into Mixed Integer Linear Programming (MILP) problems. AccMILP considers the impact of individual neurons on target labels in DNNs and combines various relaxation methods to reduce the size of NNV models while ensuring verification accuracy. The experimental results indicate that AccMILP can reduce the size of the verification model by approximately 30% and decrease the solution time by at least 80% while maintaining performance equal to or greater than 60% of MIPVerify. In other words, AccMILP is well-suited for the verification of large-scale DNNs.
The integration of artificial intelligence in agriculture has revolutionized farming practices, enhancing crop yields and resource efficiency. However, existing machine learning systems primarily focus on livestock, o...
详细信息
In the field of software development, ensuring the accuracy and quality of code remains a paramount concern. The task of precisely classifying code as correct or incorrect poses inherent challenges. This research intr...
详细信息
This paper examines the blessings of dynamic communications software program design in networks. By leveraging advances in generation, dynamic communications provide an ever-evolving method of connecting, communicatin...
详细信息
This paper examines the adoption of Problem-Based Learning (PBL) in the degree of computerengineering at the Faculty of engineering. The study presents the degree structure and the curriculum that integrates competen...
详细信息
The rapid advancement of high-quality image generation models based on AI has generated a deluge of anime illustrations. Recommending illustrations to users has become a challenge. However, existing anime recommendati...
详细信息
ISBN:
(纸本)9789819755547;9789819755554
The rapid advancement of high-quality image generation models based on AI has generated a deluge of anime illustrations. Recommending illustrations to users has become a challenge. However, existing anime recommendation systems (RS) have focused on text features but still need to integrate image features. In addition, most multi-modal (MM) RS research is constrained by tightly coupled datasets, limiting its applicability to illustrations RS. We propose the User-aware Multimodal Animation Illustration Recommendation Fusion with Painting Style (UMAIR-FPS) to tackle these gaps. In the feature extract phase, for image features, we are the first to combine painting style with semantic features to construct a dual-output image encoder for enhancing representation. For text features, we obtain embeddings based on fine-tuning Sentence-Transformers by incorporating domain knowledge that composes a variety of anime text pairs from multilingual mappings, entity relationships, and term explanation perspectives, respectively. In the MM fusion phase, we novelly propose a user-aware multi-modal contribution measurement mechanism to weight MM features dynamically according to user features at the interaction level and employ the DCN-V2 module to model bounded-degree MM crosses effectively. UMAIR-FPS surpasses the SOTA baselines on large real-world datasets.
This research discusses the method of dataset collection automatization for microwave filter synthesis by integrating machine learning techniques, thus reducing development time. Utilizing the 3D electromagnetic analy...
详细信息
In the transition to the"digital" era, many large enterprises have adopted enterprise architecture methodologies for top-level design, business architecture as the core content of enterprise architecture, it...
详细信息
Predicting water quality is essential to preserving human health and environmental sustainability. Traditional water quality assessment methods often face scalability and real-time monitoring limitations. With accurac...
详细信息
暂无评论