Siddhant Saxena
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Siddhant Saxena has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Amit Kumar Bairwa
Department of Artificial Intelligence and Machine Learning, School of Computer Science and Engineering, Manipal University Jaipur, Jaipur, India
1 shared publication - Ashish Kumar Srivastava
Manipal University Jaipur
1 shared publication - Debashis Mukherjee
Meesho, Bengaluru, India
1 shared publication - Devershi Pallavi Bhatt
Department of Computer Applications, Manipal University Jaipur, Jaipur, India
1 shared publication - Dhaneshwar Mishra
Department of Mechanical Engineering, Manipal University Jaipur, Jaipur, India
1 shared publication - Dinesh Kumar Saini
Manipal University Jaipur
1 shared publication
Latest Publications
- Open Access
- 24 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 2548-2559
DPHGNN: A Dual Perspective Hypergraph Neural Networks
Message passing on hypergraphs has been a standard framework for learning higher-order correlations between hypernodes. Recently-proposed hypergraph neural networks (HGNNs) can be categorized into spatial and spectral methods based on their design choices. In this work, we analyze the impact of change in hypergraph topology on the suboptimal performance of HGNNs and propose DPHGNN, a novel dual-perspective HGNN that introduces equivariant operator learning to capture lower-order semantics by inducing topology-aware spatial and spectral inductive biases. DPHGNN employs a unified framework to dynamically fuse lower-order explicit feature representations from the underlying graph into the super-imposed hypergraph structure. We benchmark DPHGNN over eight benchmark hypergraph datasets for the semi-supervised hypernode classification task and obtain superior performance compared to seven state-of-the-art baselines. We also provide a theoretical framework and a synthetic hypergraph isomorphism test to express the power of spatial HGNNs and quantify the expressivity of DPHGNN beyond the Generalized Weisfeiler Leman (1-GWL) test. Finally, DPHGNN was deployed by our partner e-commerce company, Meesho for the Return-to-Origin (RTO) prediction task, which shows ~7% higher macro F1-Score than the best baseline.
- 24 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 2548-2559
Open AccessDPHGNN: A Dual Perspective Hypergraph Neural Networks
Message passing on hypergraphs has been a standard framework for learning higher-order correlations between hypernodes. Recently-proposed hypergraph neural networks (HGNNs) can be categorized into spatial and spectral methods based on their design choices. In this work, we analyze the impact of change in hypergraph topology on the suboptimal performance of HGNNs and propose DPHGNN, a novel dual-perspective HGNN that introduces equivariant operator learning to capture lower-order semantics by inducing topology-aware spatial and spectral inductive biases. DPHGNN employs a unified framework to dynamically fuse lower-order explicit feature representations from the underlying graph into the super-imposed hypergraph structure. We benchmark DPHGNN over eight benchmark hypergraph datasets for the semi-supervised hypernode classification task and obtain superior performance compared to seven state-of-the-art baselines. We also provide a theoretical framework and a synthetic hypergraph isomorphism test to express the power of spatial HGNNs and quantify the expressivity of DPHGNN beyond the Generalized Weisfeiler Leman (1-GWL) test. Finally, DPHGNN was deployed by our partner e-commerce company, Meesho for the Return-to-Origin (RTO) prediction task, which shows ~7% higher macro F1-Score than the best baseline.
- 22 July 2022
- Published by Taylor & Francis
- p. 185-200
Mixed Deep Learning and Statistical Approach to Network Anomaly Detection
Anomaly detection is the identification of suspicious data patterns in the normal flow of a network that differ from the majority of data. Detection can help to prevent major attacks on an organization. However, it is difficult to prepare a model that performs well in a zero-day attack situation because obtaining the required anomaly-containing data to train the model is complex, as both attacks and anomalies vary according to the situation. The deep learning model-based approach taps the network to capture freely available data, acquiring a .pcap file from which features can be extracted using a Python-based script. The Gaussian distribution probabilities of each packet can then be computed because an anomaly-containing packet will deviate from the normal distribution. The relevant features and their mapping values are extracted from raw packets and captured in a CSV file format to act as an input tensor for the deep learning model. The output of the model gives the probabilities of normal and anomalous behaviour of the packet. The accuracy score is calculated using a statistical joint probability-based approach, which assumes all the features are independent, and the anomaly is verified by comparing the probabilities to a specific threshold.
- 22 July 2022
- Published by Taylor & Francis
- p. 185-200
Mixed Deep Learning and Statistical Approach to Network Anomaly Detection
Anomaly detection is the identification of suspicious data patterns in the normal flow of a network that differ from the majority of data. Detection can help to prevent major attacks on an organization. However, it is difficult to prepare a model that performs well in a zero-day attack situation because obtaining the required anomaly-containing data to train the model is complex, as both attacks and anomalies vary according to the situation. The deep learning model-based approach taps the network to capture freely available data, acquiring a .pcap file from which features can be extracted using a Python-based script. The Gaussian distribution probabilities of each packet can then be computed because an anomaly-containing packet will deviate from the normal distribution. The relevant features and their mapping values are extracted from raw packets and captured in a CSV file format to act as an input tensor for the deep learning model. The output of the model gives the probabilities of normal and anomalous behaviour of the packet. The accuracy score is calculated using a statistical joint probability-based approach, which assumes all the features are independent, and the anomaly is verified by comparing the probabilities to a specific threshold.
- RetractedOpen Access
- 8 June 2022
- Published by Hindawi Limited in Computational Intelligence and Neuroscience
- Vol. 2022, 1-14
GRUBin: Time-Series Forecasting-Based Efficient Garbage Monitoring and Management System for Smart Cities
The waste management of an evolving smart city environment is one of the most important tasks as the living conditions and health of the population depend on proper waste management. Currently deployed systems are failing to monitor the garbage production as they use IoT-based pipelines to monitor the production in a locality, but often the device is used to get destroyed by the frequent use of dustbin. This leads to an increase in expenditure and affects the sustainability of the system. In this work, we propose an efficient and scalable garbage monitoring and collection methodology based on time-series forecasting techniques. The proposed system is also cost-effective because of the iterative deployment of rented IoT sensors, which are used to collect time series format data and then used to train the forecasting module to learn the temporal representation of the data that can produce accurate results for monitoring the fill-up time of the garbage collector. We also propose an efficient collection in-routing technique based on the ranking of bin stations on the basis of temporal and spatial data of the fill-up time and route location to minimize the collection time by making an efficient routing algorithm for garbage collection. This concept of garbage collection will be very useful for smart city planners.
- 8 June 2022
- Published by Hindawi Limited in Computational Intelligence and Neuroscience
- Vol. 2022, 1-14
RetractedOpen AccessGRUBin: Time-Series Forecasting-Based Efficient Garbage Monitoring and Management System for Smart Cities
The waste management of an evolving smart city environment is one of the most important tasks as the living conditions and health of the population depend on proper waste management. Currently deployed systems are failing to monitor the garbage production as they use IoT-based pipelines to monitor the production in a locality, but often the device is used to get destroyed by the frequent use of dustbin. This leads to an increase in expenditure and affects the sustainability of the system. In this work, we propose an efficient and scalable garbage monitoring and collection methodology based on time-series forecasting techniques. The proposed system is also cost-effective because of the iterative deployment of rented IoT sensors, which are used to collect time series format data and then used to train the forecasting module to learn the temporal representation of the data that can produce accurate results for monitoring the fill-up time of the garbage collector. We also propose an efficient collection in-routing technique based on the ranking of bin stations on the basis of temporal and spatial data of the fill-up time and route location to minimize the collection time by making an efficient routing algorithm for garbage collection. This concept of garbage collection will be very useful for smart city planners.