Priyadarshini Kumari

prof_pic.jpg

I work at Apple, focusing on the intersection of machine learning and health.

Previously at Sony AI, I contributed to a range of projects, from developing data-efficient machine learning techniques for graph neural networks to creating multimodal perception models integrating text, and olfactory inputs. My work spanned applications in biomedical research, olfaction, and gastronomy.

I received my Ph.D. from IIT Bombay, advised by Prof. Subhasis Chaudhuri and Prof. Siddhartha Chaudhuri. My thesis was on Label-Efficient Distance Metric Learning. Before that, I completed my master’s also from IIT Bombay where I developed multimodal rendering techniques that combined haptic, visual, and auditory feedback to make 3D models of heritage sites accessible to the visually impaired.

Here is my CV

news

Jul 25, 2024 Our paper “Link prediction for hypothesis generation: an active curriculum learning infused temporal graph-based approach” is accepted at Artificial Intelligence Review 2024.
Jul 12, 2024 Our paper “CosFairNet:A Parameter-Space based Approach for Bias Free Learning” is accepted at BMVC 2024.
Oct 31, 2023 Our paper “FRUNI and FTREE synthetic knowledge graphs for evaluating explainability” is accepted at NeurIPS XAIA workshop 2023.
Sep 21, 2023 Two papers “Perceptual metrics for odorants: learning from non-expert similarity feedback using machine learning” and “Comparing molecular representations, e-nose signals, and other featurization, for learning to smell aroma molecules” are accepted at PLOS One
Jul 15, 2023 Our paper “Optimizing Learning Across Multimodal Transfer Features for Modeling Olfactory Perception” was accepted to Multimodal SIGKDD 2023

selected publications

  1. Link prediction for hypothesis generation: an active curriculum learning infused temporal graph-based approach
    Uchenna Akujuobi* , Priyadarshini Kumari, Jihun Choi , Samy Badreddine , Kana Maruyama , and 2 more authors
    In Artificial Intelligence Review , 2024

    We present THiGER-A, a solution aimed at addressing the hypothesis-generation problem. This method utilizes a temporal graph-based node-pair embedding and incorporates an active-curriculum training approach to precisely capture the dynamic evolution of discoveries over time.

  2. FRUNI and FTREE synthetic knowledge graphs for evaluating explainability
    Pablo Sanchez Martin , Tarek Besold , and Priyadarshini Kumari
    In NeurIPS 2023 Workshop XAIA , 2023

    We introduce two synthetic datasets, FRUNI and FTREE, to assess explainer methods’ ability to identify predictions relying on indirectly connected links.

  3. Optimizing Learning Across Multimodal Transfer Features for Modeling Olfactory Perception
    Daniel Shin , Gao Pei , Priyadarshini Kumari, and Tarek Besold
    In International Workshop on Multimodal Learning at SIGKDD , 2023

    We introduce a novel multilabel and multimodal transfer learning technique for modeling olfactory perception. Our approach aims to tackle the challenges of data scarcity and label skewness in the olfactory domain.

  4. A Unified Batch Selection Policy for Active Metric Learning
    Priyadarshini Kumari, Siddhartha Chaudhuri , Vivek Borkar , and Subhasis Chaudhuri
    In ECML-PKDD , 2021

    We propose a batch-mode active learning method that balances informativeness and diversity of batches of triplets combinedly.

  5. Batch decorrelation for active metric learning
    Priyadarshini Kumari, Ritesh Goru , Siddhartha Chaudhuri , and Subhasis Chaudhuri
    In IJCAI , 2021

    We propose triplet-based decorrelation measures to improve the performance of batch-mode active metric learning strategies.

  6. PerceptNet: Learning perceptual similarity of haptic textures in presence of unorderable triplets
    Priyadarshini Kumari, Siddhartha Chaudhuri , and Subhasis Chaudhuri
    In IEEE World Haptics Conference (WHC) , 2019

    We propose a deep metric learning approach that demonstrates the utility of ambiguous triplets for effectively modeling human perception.