Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: To what extent does the noise injection strategy in XSimGCL improve downstream recommendation performance (NDCG@10) when pre-trained on corrupted user-item graphs compared to baseline augmentation. Human…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: What is the impact of different negative sampling strategies on the robustness of self-supervised recommendation models like LightGCL, SimGCL, and DCL when evaluated on sparse HOI datasets using. In the last few…

Abstract: This report synthesises findings from 7 peer-reviewed papers addressing the following research question: How does the performance of simplified noise injection in graph contrastive learning compare to heavy augmentation techniques in terms of mean average precision (MAP) on extreme sparse recommendation. Abstract…

Abstract: This report synthesises findings from 7 peer-reviewed papers addressing the following research question: How do LightGCL, SimGCL, and DCL scale in terms of model performance (measured by AUC and NDCG) when applied to cross-domain recommendation tasks with varying dataset sizes and sparsity levels. Abstract The advent…

Abstract: This report synthesises findings from 2 peer-reviewed papers addressing the following research question: How does the inference efficiency of LightGCL compare to SimGCL and DCL when evaluated on perturbed HOI datasets with varying levels of data sparsity, measured by throughput and latency. Edge computing has become…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: Can multimodal contrastive learning models outperform unimodal counterparts in cross-domain recommendation tasks, as measured by mAP@k and HR@k metrics on benchmark datasets like MovieLens and Last.fm. Unlike…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: How does adversarial training affect the token generation throughput and latency of large language models on the HumanEval code generation benchmark. This article presents a comprehensive and practical guide for…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: Do contrastive learning frameworks such as LightGCL and SimGCL demonstrate improved robustness to noisy data when evaluated using recall@k and NDCG@k metrics on corrupted recommendation datasets. Multilayer…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How do multimodal similarity measures (e.g., CLIP embeddings) compare to text-only measures (e.g., SimGCL, DCL) in clustering web-page content that includes both text and images, as measured by. We propose a dual…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: What is the performance trade-off between inference speed and clustering accuracy when comparing lightweight similarity measures versus computationally intensive ones (e.g., BERTScore vs. TF-IDF) on. This paper…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: How do LightGCL and SimGCL differ in their robustness to noisy user-item interactions when evaluated using Recall@K and NDCG@K on extremely sparse benchmark datasets. Graph neural network (GNN) is a powerful…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the choice of similarity measure (e.g., cosine, Jaccard, Euclidean) impact the cluster quality of transformer-based document embeddings (e.g., BERT, RoBERTa) when evaluated using adjusted. This paper…

Abstract: This report synthesises findings from 7 peer-reviewed papers addressing the following research question: What is the impact of pre-training LightGCL on large-scale graphs on its performance in Hit Ratio@5 and NDCG@10 compared to other contrastive learning methods like GraCL and MVGRL. Graph neural network (GNN) is a…

Abstract: This report synthesises findings from 7 peer-reviewed papers addressing the following research question: How robust is XSimGCL's dual-target recommendation capability when evaluated on out-of-distribution datasets using alignment preservation and recommendation diversity metrics. Self-supervised learning (SSL) has…

Abstract: This report synthesises findings from 8 peer-reviewed papers addressing the following research question: How does XSimGCL's domain adaptation performance compare to baseline models on cross-domain recommendation tasks when evaluated using NDCG@10 and accuracy metrics across different domain pairs. Recommender system…

Abstract: This report synthesises findings from 2 peer-reviewed papers addressing the following research question: How does the inference efficiency of LightGCL compare to SGL and GCA when scaling to large-scale recommendation datasets like Amazon-1M and MovieLens-20M in terms of runtime and memory usage. In recent years, deep…

Abstract: This report synthesises findings from 9 peer-reviewed papers addressing the following research question: How does the robustness of LightGCL compare to GraCL and MVGRL when evaluated on Hit Ratio@5 and NDCG@10 under adversarial graph perturbations in dense interaction graphs. Point clouds provide a flexible geometric…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: What is the relative robustness of LightGCL versus SGL and GCA under varying levels of data sparsity in user-item interactions, as measured by precision@10 and recall@20 on synthetic. Deep convolutional neural…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: How does the performance of LightGCL, SGL, and GCA vary when applied to cross-domain recommendation tasks, such as transferring from MovieLens-100K to Amazon Book Reviews, in terms of precision@10. In recent…

Abstract: This report synthesises findings from 1 peer-reviewed paper addressing the following research question: How do multimodal reasoning benchmarks compare the adversarial robustness of graph contrastive learning and vision-language contrastive models when evaluated under varying levels of input. Deep learning has shown…

Abstract: This report synthesises findings from 16 peer-reviewed papers addressing the following research question: What is the relative drop in node classification accuracy for graph diffusion models versus message-passing GNNs when evaluated on large-scale graphs subjected to increasing levels of spectral. Graph Neural…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: What is the trade-off between inference efficiency (e.g., throughput in tokens/sec) and model robustness (e.g., accuracy on noisy CIFAR-10-C) when applying gradient-based sparsification to. Abstract Realizing…

Abstract: This report synthesises findings from 9 peer-reviewed papers addressing the following research question: How does the memory complexity scaling of graph diffusion models during inference compare to sparse GNN architectures when processing large graphs with high-frequency spectral perturbations. Abstract Deep learning…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does the inference throughput of graph diffusion models compare to traditional GNNs when scaling to graphs with over 100,000 nodes under adversarial spectral perturbations. Machine learning plays an…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: To what extent does the sparsification level (e.g., 50\% vs. 90\% edge reduction) affect the robustness of contrastive learning-based recommenders against adversarial attacks, as measured by metrics.…