Deep learning-based algorithms, e.g., convolutional networks, have significantly facilitated multivariate time series classification (MTSC) task. Nevertheless, they suffer from the limitation in modeling long-range dependence due to nature of convolution operations. Recent advancements shown potential transformers capture dependence. However, it would incur severe issues, such as fixed scale representations, temporal-invariant and quadratic complexity, with directly applicable MTSC task...

Enhancing the expressive capacity of deep learning-based time series models with self-supervised pre-training has become ever-increasingly prevalent in classification. Even though numerous efforts have been devoted to developing for data, we argue that current methods are not sufficient learn optimal representations due solely unidirectional encoding over sparse point-wise input units. In this work, propose TimeMAE, a novel paradigm learning transferrable based on transformer networks. The...

Recent years have witnessed great success in deep learning-based sequential recommendation (SR), which can provide more timely and accurate recommendations. One of the most effective SR architectures is to stack high-performance residual blocks, e.g., prevalent self-attentive convolutional operations, for capturing long- short-range dependence behaviors. By carefully revisiting previous models, we observe: 1) simple architecture modification gating each connection help us train deeper models...

The reasoning abilities are one of the most enigmatic and captivating aspects large language models (LLMs). Numerous studies dedicated to exploring expanding boundaries this capability. However, tasks that embody both recall characteristics often overlooked. In paper, we introduce such a novel task, code reasoning, provide new perspective for LLMs. We summarize three meta-benchmarks based on established forms logical instantiate these into eight specific benchmark tasks. Our testing...

Recent efforts have been dedicated to enhancing time series forecasting accuracy by introducing advanced network architectures and self-supervised pretraining strategies. Nevertheless, existing approaches still exhibit two critical drawbacks. Firstly, these methods often rely on a single dataset for training, limiting the model's generalizability due restricted scale of training data. Secondly, one-step generation schema is widely followed, which necessitates customized head overlooks...

Recent efforts have been dedicated to enhancing time series forecasting accuracy by introducing advanced network architectures and self-supervised pretraining strategies. Nevertheless, existing approaches still exhibit two critical drawbacks. Firstly, these methods often rely on a single dataset for training, limiting the model's generalizability due restricted scale of training data. Secondly, one-step generation schema is widely followed, which necessitates customized head overlooks...

Deep learning has brought significant breakthroughs in sequential recommendation (SR) for capturing dynamic user interests. A series of recent research revealed that models with more parameters usually achieve optimal performance SR tasks, inevitably resulting great challenges deploying them real systems. Following the simple assumption light networks might already suffice certain users, this work, we propose CANet, a conceptually yet very scalable framework assigning adaptive network...

The fundamental task of intelligent educational systems is to offer adaptive learning services students, such as exercise recommendations and computerized testing. However, optimizing required models in these would always encounter the collection difficulty high-quality interaction data practice. Therefore, establishing a student simulator great value since it can generate valid interactions help optimize models. Existing advances have achieved success but generally suffer from exposure bias...

Deep learning-based algorithms, e.g., convolutional networks, have significantly facilitated multivariate time series classification (MTSC) task. Nevertheless, they suffer from the limitation in modeling long-range dependence due to nature of convolution operations. Recent advancements shown potential transformers capture dependence. However, it would incur severe issues, such as fixed scale representations, temporal-invariant and quadratic complexity, with directly applicable MTSC task...

For the advancements of time series classification, scrutinizing previous studies, most existing methods adopt a common learning-to-classify paradigm - classifier model tries to learn relation between sequence inputs and target label encoded by one-hot distribution. Although effective, this conceals two inherent limitations: (1) encoding categories with distribution fails reflect comparability similarity labels, (2) it is very difficult transferable across domains, which greatly hinder...

Self-supervised learning has become a popular and effective approach for enhancing time series forecasting, enabling models to learn universal representations from unlabeled data. However, effectively capturing both the global sequence dependence local detail features within data remains challenging. To address this, we propose novel generative self-supervised method called TimeDART, denoting Diffusion Auto-regressive Transformer Time forecasting. In treat patches as basic modeling units....

Time series forecasting is vital in numerous web applications, influencing critical decision-making across industries. While diffusion models have recently gained increasing popularity for this task, we argue they suffer from a significant drawback: indiscriminate noise addition to the original time followed by denoising, which can obscure underlying dynamic evolving trend and complicate forecasting. To address limitation, propose novel flexible decoupled framework (FDF) that learns...

Intraoperative hypotension (IOH) prediction using Mean Arterial Pressure (MAP) is a critical research area with significant implications for patient outcomes during surgery. However, existing approaches predominantly employ static modeling paradigms that overlook the dynamic nature of physiological signals. In this paper, we introduce novel Hybrid Multi-Factor (HMF) framework reformulates IOH as blood pressure forecasting task. Our leverages Transformer encoder, specifically designed to...

Sequential recommendation models user interests based on historical behaviors to provide personalized recommendation. Previous sequential algorithms primarily employ neural networks extract features of interests, achieving good performance. However, due the system datasets sparsity, these often small-scale network frameworks, resulting in weaker generalization capability. Recently, a series large pre-trained language have been proposed. Nonetheless, given real-time demands systems, challenge...

Multivariate time series forecasting plays a crucial role in various real-world applications. Significant efforts have been made to integrate advanced network architectures and training strategies that enhance the capture of temporal dependencies, thereby improving accuracy. On other hand, mainstream approaches typically utilize single unified model with simplistic channel-mixing embedding or cross-channel attention operations account for critical intricate inter-channel dependencies....

Sequential recommender systems (SRS) have gained widespread popularity in recommendation due to their ability effectively capture dynamic user preferences. One default setting the current SRS is uniformly consider each historical behavior as a positive interaction. Actually, this has potential yield sub-optimal performance, item makes distinct contribution user's interest. For example, purchased items should be given more importance than clicked ones. Hence, we propose general automatic...

Deep learning has brought significant breakthroughs in sequential recommendation (SR) for capturing dynamic user interests. A series of recent research revealed that models with more parameters usually achieve optimal performance SR tasks, inevitably resulting great challenges deploying them real systems. Following the simple assumption light networks might already suffice certain users, this work, we propose CANet, a conceptually yet very scalable framework assigning adaptive network...