Kamal Berahmand, Ph.D.
Dr. Kamal Berahmand received his Ph.D. in Computer Science from Queensland University of Technology (QUT), Brisbane, in early 2025, where he specialized in graph-based machine learning and representation learning. He is currently a Postdoctoral Research Fellow at RMIT University, Melbourne, Australia. He works on key tasks including clustering, link prediction, node classification, and graph anomaly detection. Currently, he is involved in two projects at RMIT: (1) investigating the impact of higher-order structures in traffic networks, and (2) developing scalable methods for graph anomaly detection.
Research Output
Dr. Kamal Berahmand has published over 50 peer-reviewed papers in the areas of machine learning, graph representation learning, and network science. His work has received significant academic attention, with 3,795 citations in total and 3,773 since 2020. He holds an h-index of 36 and an i10-index of 48. His research has been featured in prestigious journals such as IEEE Transactions on Knowledge and Data Engineering, Pattern Recognition, ACM Computing Surveys, IEEE Transactions on Computational Social Systems, and IEEE Transactions on Network Science and Engineering.
Graph Learning
Dr. Berahmand’s research in graph learning centers around three core areas:
🔹 Graph Embedding
Learns low-dimensional representations of nodes and graphs using models such as GCNs, GATs, graph autoencoders, and graph recurrent networks, enabling accurate performance in tasks like classification and link prediction.
🔹 Graph Reduction
Improves scalability and training efficiency in large-scale GNNs using graph modification techniques such as coarsening, sparsification, and condensation, along with inference optimization methods including pruning, quantization, and distillation.
🔹 Graph Structure Learning
Constructs optimal graph topologies directly from raw data using pairwise, anchor-based, and hypergraph methods in both fixed and adaptive settings, enhancing the quality of learned representations.
