SIGNAL ANALYSIS FOR LARGE SCALE APPLICATIONS
Next Generation Deep Learning
We are out to replace several decades old backpropagation with state-of-the-art optimizing paradigms, bringing in new capabilities in deep learning, such as learning unsupervised convolutional filters and marrying two of the hottest topics in machine learning and signal processing – deep learning with graph signal processing.
- Shalini Sharma and Angshul Majumdar, "A Deep State Space Model for Predicting Cryptocurrency Price", Information Sciences 2023.
- Anurag Goel, Angshul Majumdar, Emilie Chouzenoux and Giovanni Chierchia, "Deep Convolutional K-Means Clustering", ICIP 2022.
Energy Analytics
Smartgrids have bought IoT and AI to the power sector. Here we develop latest machine learning models that are tailored for this industry. The emphasis is primarily AI based demand side management with human in the loop. We concentrate on energy disaggregation, demand forecasting and anomaly detection.
- Shikha Singh, Angshul Majumdar, Emilie Chouzenoux and Giovanni Chierchia, "Multi-label Deep Convolutional Transform Learning for Non-intrusive Load Monitoring", ACM Transactions on Knowledge Discovery from Data 2021.
- Shalini Sharma and Angshul Majumdar, "Unsupervised Detection of Non-Technical Losses via Recursive Transform Learning", IEEE Transactions on Power Delivery 2021.
Bioinformatics
In the past, we have worked on analysis of single cell RNA sequences. Currently our efforts are geared towards problems on computational drug discovery, e.g. drug target interactions, drug disease association and drug virus association.
- Pooja Gupta, Angshul Majumdar, Emilie Chouzenoux and Giovanni Chierchia, "DeConDFFuse: Predicting Drug-Drug Interaction using joint Deep Convolutional Transform Learning and Decision Forest fusion framework", Expert Systems With Applications 2023.
- A. Mongia, S. Jain, E. Chouzenoux and A. Majumdar, "DeepVir - Graphical Deep Matrix Factorization for In Silico Antiviral Repositioning: Application to COVID-19", Journal of Computational Biology 2022.
Remote Sensing
We develop novel machine learning algorithms for hyperspectral image classification. The emphasis is on inferencing in low-data regimes.
- V. Singhal and A. Majumdar, “Row-Sparse Discriminative Deep Dictionary Learning for Hyperspectral Image Classification”, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol. 11 (12), 5019 – 5028, 2019.
- V. Singhal, H. Agrawal, S. Tariyal and A. Majumdar, “Discriminative Robust Deep Dictionary Learning for Hyperspectral Image Classification”, IEEE Transactions on Geosciences and Remote Sensing, Vol. 55 (9), pp. 5274-5283, 2017.
Inverse Problems
Solution of an under-determined system of linear equations finds many applications in signal processing. Tasks like reconstruction, denoising, deblurring, deconvolution, inverse half-toning, super-resolution, all fall under the category of inverse problems. We solve such problems by interecting the latest in signal processing and machine learning.
- V. Singhal and A. Majumdar, “A domain adaptation approach to solve inverse problems in imaging via coupled deep dictionary learning”, Pattern Recognition, vol. 100, 2020.
- J. Mehta and A. Majumdar, “RODEO: Robust DE-aliasing autoencOder for Real-time Medical Image Reconstruction”, Pattern Recognition, Vol. 63, pp. 499-510, 2017.
Recommender Systems
Collaborative filtering is synonymous with recommender systems. We formulate it as a matrix completion problem and develop powerful mathematical tools that combines the former with graph signal processing.
- A. Mongia, N. Jhamb, E. Chouzenoux and A. Majumdar, “Deep Latent Factor Model for Collaborative Filtering”, Signal Processing, Vol. 169, 107366, 2020.
- A. Mongia and A. Majumdar, “Matrix Completion on Multiple Graphs: Application in Collaborative Filtering”, Signal Processing, Vol. 165, pp. 144-148, 2019.