My research interests include Disentangled Representation Learning, Causality, Explainability, Computational Entrepreneurship, Mathematical Modelling particularly in building end-to-end systems with high socio-technical impacts. [Curriculum Vitae] [Google Scholar]
Time-domain, frequency-domain and time-frequency representations of a sine waveform across different frequencies sampled at 16KHz for 1 seconds. The top row is the original signal and the bottom row is the reconstructed version for the disentangled Variational Auto Encoder.Time-domain, time-frequency and frequency-domain representations of a sawtooth waveform across different frequencies sampled at 16KHz for 2 seconds each.Visualizing the convergce of clusters of classes in the MNIST validation set using both Triplet (Left) and Supervised Contrastive Representation LearningVisualizing the convergence of latent representations over training iterations for modified Barlow Twins model: decorrelated audio representations (left) and original small training data (right). The 128-dimensional latent space reveals the progression of learned features and their alignment throughout the training process.A visual illustration of two divergence measure: the Kullback–Leibler (KL) and Jensen–Shannon (JS) divergences between two probability distributions $P_x $ and $Q_x$.Explainable AI in Medical Image Analysis using Visual Saliency Maps. Feature attribution was done with a trained DenseNet-121 model.Learnt 2D and 3D principal component latent representations of both human and non-human primates.Learnt 2D and 3D principal component latent representations of speeches at the United States Congress by five world leaders.
February 16, 2024; “Learning Disentangled Audio Representations through Controlled Synthesis” accepted for oral presentation at ICLR Tiny Papers 2024.
February 15, 2024; “Understanding Self-Supervised Learning of Speech Representation via Invariance and Redundancy Reduction “ accepted for oral presentation at MDPI Information 2024, 15(2), 114.
December 10, 2023; Attended NeurIPS 2023 in New Orleans, USA where Presented research poster on disentangled speech representation learning via self-supervision. Sharing novel techniques for learning interpretable and robust speech representations.
September 23, 2023; “Self-Supervised Learning of Speech Representation via Redundancy Reduction” extended abstract published at Gesellschaft für Informatik e.V..
August 1, 2023; “Visual Interpretable and Explainable Deep Learning Models for Brain Tumor MRI and COVID-19 Chest X-ray Images” published in arXiv preprint arXiv:2208.00953.
June 15, 2021; Published a dataset that was curated in collaboration between the Computer Science and Engineering Department, University of Dhaka and the National Institute of Neuroscience, Bangladesh. It comprise 5,285 T1-weighted contrast- enhanced brain MRI images belonging to 38 categories. Available here Brain MRI Dataset.
