Machine learning (ML) methods have been developing rapidly, but configuring and selecting proper to achieve a desired performance is increasingly difficult tedious. To address this challenge, automated machine (AutoML) has emerged, which aims generate satisfactory ML configurations for given tasks in data-driven way. In paper, we provide comprehensive survey on topic. We begin with the formal definition of AutoML then introduce its principles, including bi-level objective, strategy,...

Knowledge graph (KG) embedding is a fundamental problem in data mining research with many real-world applications. It aims to encode the entities and relations into low dimensional vector space, which can be used for subsequent algorithms. Negative sampling, samples negative triplets from non-observed ones training data, an important step KG embedding. Recently, generative adversarial network (GAN), has been introduced sampling. By sampling large scores, these methods avoid of vanishing...

Scoring functions (SFs), which measure the plausibility of triplets in knowledge graph (KG), have become crux KG embedding. Lots SFs, target at capturing different kinds relations KGs, been designed by humans recent years. However, as can exhibit complex patterns that are hard to infer before training, none them consistently perform better than others on existing benchmark data sets. In this paper, inspired success automated machine learning (AutoML), we propose automatically design SFs...

Reasoning on the knowledge graph (KG) aims to infer new facts from existing ones. Methods based relational path have shown strong, interpretable, and transferable reasoning ability. However, paths are naturally limited in capturing local evidence graphs. In this paper, we introduce a novel structure, i.e., directed (r-digraph), which is composed of overlapped paths, capture KG's evidence. Since r-digraphs more complex than how efficiently construct effectively learn them challenging....

Due to the popularity of Graph Neural Networks (GNNs), various GNN-based methods have been designed reason on knowledge graphs (KGs). An important design component KG reasoning is called propagation path, which contains a set involved entities in each step. Existing use hand-designed paths, ignoring correlation between and query relation. In addition, number will explosively grow at larger steps. this work, we are motivated learn an adaptive path order filter out irrelevant while preserving...

Addressing the structural and functional variability between subjects for robust affective brain-computer interface (aBCI) is challenging but of great importance, since calibration phase aBCI time-consuming. In this paper, we propose a subject transfer framework electroencephalogram (EEG)-based emotion recognition via component analysis. We compare two state-of-the-art subspace projecting approaches called analysis (TCA) kernel principle (KPCA) transfer. The main idea to learn set components...

Abstract The brute‐force scaleup of training datasets, learnable parameters and computation power, has become a prevalent strategy for developing more robust learning models. However, due to bottlenecks in data, computation, trust, the sustainability this is serious concern. In paper, we attempt address issue parsimonious manner (i.e., achieving greater potential with simpler models). key drive models using domain‐specific knowledge, such as symbols, logic, formulas, instead purely relying...

Recent advances in large language models (LLMs) have showcased exceptional performance long-context tasks, while facing significant inference efficiency challenges with limited GPU memory. Existing solutions first proposed the sliding-window approach to accumulate a set of historical \textbf{key-value} (KV) pairs for reuse, then further improvements selectively retain its subsets at each step. However, due sparse attention distribution across long context, it is hard identify and recall...

Learning embeddings for entities and relations in knowledge graph (KG) have benefited many downstream tasks. In recent years, scoring functions, the crux of KG learning, been human designed to measure plausibility triples capture different kinds KGs. However, as exhibit intricate patterns that are hard infer before training, none them consistently perform best on benchmark this paper, inspired by success automated machine learning (AutoML), we search bilinear functions tasks through AutoML...

The scoring function, which measures the plausibility of triplets in knowledge graphs (KGs), is key to ensure excellent performance KG embedding, and its design also an important problem literature. Automated machine learning (AutoML) techniques have recently been introduced into task-aware functions, achieve state-of-the-art embedding. However, effectiveness searched functions still not as good desired. In this paper, observing that existing can exhibit distinct on different semantic...

While hyper-parameters (HPs) are important for knowledge graph (KG) learning, existing methods fail to search them efficiently. To solve this problem, we first analyze the properties of different HPs and measure transfer ability from small subgraph full graph. Based on analysis, propose an efficient two-stage algorithm KGTuner, which efficiently explores HP configurations at stage transfers top-performed fine-tuning large second stage. Experiments show that our method can consistently find...

Knowledge graph (KG) embedding is well-known in learning representations of KGs. Many models have been proposed to learn the interactions between entities and relations triplets. However, long-term information among multiple triplets also important KG. In this work, based on relational paths, which are composed a sequence triplets, we define Interstellar as recurrent neural architecture search problem for short-term along paths. First, analyze difficulty using unified model work...

Knowledge graph (KG) embedding is a fundamental task in natural language processing, and various methods have been proposed to explore semantic patterns distinctive ways. In this paper, we propose learn an ensemble by leveraging existing relation-aware manner. However, exploring these semantics using leads much larger search space than general methods. To address issue, divide-search-combine algorithm RelEns-DSC that searches the relation-wise weights independently. This has same computation...