JONNEE: Joint Network Nodes and Edges Embedding

Makarov, Ilya; Korovina, Ksenia; Kiselev, Dmitrii

doi:10.1109/access.2021.3122100

Cited by 26 publications

(5 citation statements)

References 63 publications

(75 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…CD show competitive and superior performance for each parser. The reason that the performance of CD is superior to SD is because it has more finegrained relation types than SD [17] . The passive construction's subject in SD is marked as a particular nsubj dependency label, while the active constructions SUB(J) is also marked as the same subject label in CD.…”

Section: Experimental Results and Analysismentioning

confidence: 99%

“…The model redefine neighbors from the perspectives of node and edge respectively, treat edges or nodes as nodes in the new graph, and learn the edge and node embedding in the new graph. Makarov et al [17] proposed a JONNEE model to combine edges and nodes embedding, which can make its edge-to-vertex interactive representation as a line graph. Some Researches discovered that the parsers with domain models using different formats can improve the event extraction system.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Child-Sum (N2e2n)Tree-Lstms: An Interactive Child-Sum Tree-Lstms to Extract Biomedical Event

Wang,

Cao,

Yuan

2023

Preprint

View full text Add to dashboard Cite

Section: Experimental Results and Analysismentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Child-Sum (N2e2n)Tree-Lstms: An Interactive Child-Sum Tree-Lstms to Extract Biomedical Event

Wang,

Cao,

Yuan

2023

Preprint

View full text Add to dashboard Cite

“…Contrastive methods involve contrasting two different views generated from graph augmentations [37]. General data augmentation techniques in graph contrastive learning (GraphCL), such as node dropping, edge permuting, and subgraph extracting, can significantly alter the chemical properties of molecules, leading to limited improvements or negative transfer on downstream tasks [38].…”

Section: Self-supervised Learning On Graphsmentioning

confidence: 99%

Predicting Molecule Toxicity via Descriptor-Based Graph Self-Supervised Learning

Li,

Makarov,

Kiselev

2023

IEEE Access

Self Cite

View full text Add to dashboard Cite

Predicting molecular properties with Graph Neural Networks (GNNs) has recently drawn a lot of attention, with compound toxicity prediction being one of the biggest challenges. In cases where there is insufficient labeled molecule data, an effective approach is to pre-train GNNs on large-scale unlabeled molecular data and then fine-tune them for downstream tasks. Among pre-training strategies, node-level pre-training involves masking and predicting atom properties, while motif-based methods capture rich information in subgraphs. These approaches have shown effectiveness across various downstream tasks. However, current pre-training frameworks face two main challenges: (1) node-level auxiliary tasks do not preserve useful domain knowledge, and (2) the fusion of motif-based methods and node-level tasks is computationally extensive. To address these challenges, we propose Descriptor-based Graph Self-supervised Learning (DGSSL), a method that utilizes domain knowledge to enhance graph representation learning. Specifically, it identifies descriptor centers in molecules and encodes motif-like information as special atomic numbers in the pre-training tasks. This enables node-level self-supervised pre-training frameworks for GNNs to also capture rich information in local subgraphs. Experimental results demonstrate that our method achieves state-of-the-art performance on three toxicity-related benchmarks.

show abstract

“…Such models are said to have limited generalization capabilities and inherent biases [12], [31]. However, there are promising avenues for improvement, such as better embedding learning [31]- [33] or the application of so-called attribute attention mechanisms [34].…”

Section: Zero-shot Learningmentioning

confidence: 99%

Scalable Zero-Shot Logo Recognition

Shulgin,

Makarov

2023

IEEE Access

Self Cite

View full text Add to dashboard Cite

Brand logo recognition is a task focused on the identification and classification of logos, with various applications such as brand protection and market discovery. Since the number of brands is dynamic and evolves over time, we propose a two-step, zero-shot framework for this problem. In the first step, we train a Scaled-YOLOv4 single-stage universal logo detector to identify regions containing logolike objects. Our results indicate that this detector achieves generalizability comparable to the two-stage Faster-RCNN model. In the second step, we employ an enhanced CLIP model for zero-shot classification of the identified regions. Our experiments demonstrate that the CLIP model outperforms state-of-the-art fewshot classifiers in terms of accuracy. Additionally, we adopt test-time augmentation to improve the model's resistance against false positives. We also present a proof-of-concept for fine-tuning the CLIP model, which enhances its cosine similarity measures. Our proposed end-to-end solution is scalable in terms of the number of brands and requires only the brand names for detection. The logo detector achieves superior performance on the FlickrLogos-32 dataset without the need for additional fine-tuning.INDEX TERMS universal logo detector, few-shot learning, logo recognition, Flickr-32

show abstract

JONNEE: Joint Network Nodes and Edges Embedding

Cited by 26 publications

References 63 publications

Child-Sum (N2e2n)Tree-Lstms: An Interactive Child-Sum Tree-Lstms to Extract Biomedical Event

Child-Sum (N2e2n)Tree-Lstms: An Interactive Child-Sum Tree-Lstms to Extract Biomedical Event

Predicting Molecule Toxicity via Descriptor-Based Graph Self-Supervised Learning

Scalable Zero-Shot Logo Recognition

Contact Info

Product

Resources

About