arXiv subjects

Paribesh Regmi

Publications and source records attributed to Paribesh Regmi.

3 recordsLinked to original sources

arXiv2026

Adaptive Two-Stage Visual Token Pruning for Efficient Inference in Video-Language Models

Vision-language models excel at image and video understanding but suffer from high inference latency due to the need to process thousands of tokens per image, limiting their deployment on resource-constrained edge devices and in real-time surveillance applications. This challenge is further amplified in video processing, where multiple frames must be analyzed simultaneously. Existing token reduction techniques are largely developed for single-image inputs and therefore fail to account for the temporal and inter-frame redundancies present in video sequences. In addition, these methods generally rely on a fixed, uniform pruning ratio applied across all inputs, which is suboptimal because the degree of redundancy can vary significantly between different videos, necessitating content-dependent pruning levels to preserve critical information. To address these limitations, we propose a two-stage adaptive token pruning strategy specifically designed for video processing. In the first stage, we prune out the redundant frames, and in the second stage, token-level pruning is applied within the retained frames. Crucially, the pruning ratio in the second stage is determined adaptively based on the content of each video. This is achieved by analyzing the correlation structure of token embeddings to quantify redundancy, which is used to determine the ratio. Importantly, our method is entirely post-hoc and requires no additional training or fine-tuning, while achieving strong empirical gains; notably, it improves accuracy by +7\% on a video captioning benchmark at 10\% token retention, while reducing computation TFLOPs by 95\%.

cs.CV↗

arXiv2026

Bayesian Neighborhood Adaptation for Graph Neural Networks

The neighborhood scope (i.e., number of hops) where graph neural networks (GNNs) aggregate information to characterize a node's statistical property is critical to GNNs' performance. Two-stage approaches, training and validating GNNs for every pre-specified neighborhood scope to search for the best setting, is a time-consuming task and tends to be biased due to the search space design. How to adaptively determine proper neighborhood scopes for the aggregation process for both homophilic and heterophilic graphs remains largely unexplored. We thus propose to model the GNNs' message-passing behavior on a graph as a stochastic process by treating the number of hops as a beta process. This Bayesian framework allows us to infer the most plausible neighborhood scope for message aggregation simultaneously with the optimization of GNN parameters. Our theoretical analysis shows that the scope inference improves the expressivity of a GNN. Experiments on benchmark homophilic and heterophilic datasets show that the proposed method is compatible with state-of-the-art GNN variants, achieving competitive or superior performance on the node classification task, and providing well-calibrated predictions. Implementation is available at : https://github.com/paribeshregmi/BNA-GNN

cs.LG↗

arXiv2022

Predicting Biomedical Interactions with Probabilistic Model Selection for Graph Neural Networks

Heterogeneous molecular entities and their interactions, commonly depicted as a network, are crucial for advancing our systems-level understanding of biology. With recent advancements in high-throughput data generation and a significant improvement in computational power, graph neural networks (GNNs) have demonstrated their effectiveness in predicting biomedical interactions. Since GNNs follow a neighborhood aggregation scheme, the number of graph convolution (GC) layers (i.e., depth) determines the neighborhood orders from which they can aggregate information, thereby significantly impacting the model's performance. However, it often relies on heuristics or extensive experimentation to determine an appropriate GNN depth for a given biomedical network. These methods can be unreliable or result in expensive computational overhead. Moreover, GNNs with more GC layers tend to exhibit poor calibration, leading to high confidence in incorrect predictions. To address these challenges, we propose a Bayesian model selection framework to jointly infer the most plausible number of GC layers supported by the data, apply dropout regularization, and learn network parameters. Experiments on four biomedical interaction datasets demonstrate that our method achieves superior performance over competing methods, providing well-calibrated predictions by allowing GNNs to adapt their depths to accommodate interaction information from various biomedical networks. Source code and data is available at: https://github.com/kckishan/BBGCN-LP/tree/master

q-bio.QM↗