Research
My research focuses on developing trustworthy and interpretable probabilistic AI systems that can reason under uncertainty, generalize reliably, and align with human decision-making.
Probabilistic Generative Models
Developing expressive and tractable generative models that can learn complex data distributions and reason probabilistically about the learned distribution, to make robust and reliable decisions.
Related Publications
Tractable Sharpness-Aware Learning of Probabilistic Circuits
AAAI '26H Suresh, Sahil Sidheekh, S Natarajan, NC Krishnan
Autoencoding Probabilistic Circuits
TPM Workshop '25S Braun, Sahil Sidheekh, S Natarajan, A Vergari, M Mundt, K Kersting
Building Expressive and Tractable Probabilistic Generative Models: A Review
IJCAI '24Sahil Sidheekh, S Natarajan
Probabilistic Flow Circuits: Towards Unified Deep Models for Tractable Probabilistic Inference
UAI '23Sahil Sidheekh, K Kersting, S Natarajan
VQ-Flows: Vector Quantized Local Normalizing Flows
UAI '22Sahil Sidheekh, CB Dock, T Jain, R Balan, MK Singh
On Characterizing GAN Convergence Through Proximal Duality Gap
ICML '21Sahil Sidheekh, A Aimen, NC Krishnan
Neuro-Symbolic AI
Combining neural and symbolic approaches to create interpretable and robust AI systems that leverage both learning and reasoning.
Related Publications
Human-in-the-loop or AI-in-the-loop? Automate or Collaborate?
AAAI '25S Natarajan, S Mathur, Sahil Sidheekh, W Stammer, K Kersting
Credibility-aware multi-modal fusion using probabilistic circuits
AISTATS '25Sahil Sidheekh, P Tenali, S Mathur, E Blasch, K Kersting, S Natarajan
A Unified Framework for Human-Allied Learning of Probabilistic Circuits
AAAI '25Athresh Karanam, Saurabh Mathur, Sahil Sidheekh, Sriraam Natarajan
Human-Allied Relational Reinforcement Learning
ACS '25FG Darvishvand, H Shindo, Sahil Sidheekh, K Kersting, S Natarajan
On the Robustness and Reliability of Late Multi-Modal Fusion using Probabilistic Circuits
FUSION '24Sahil Sidheekh, P Tenali, S Mathur, E Blasch, S Natarajan
Credibility-aware Reliable Multi-Modal Fusion Using Probabilistic Circuits
AAAI '24S Mathur, Sahil Sidheekh, P Tenali, E Blasch, K Kersting, S Natarajan
Bayesian learning of probabilistic circuits with domain constraints
TPM Workshop '23A Karanam, S Mathur, Sahil Sidheekh, S Natarajan
Meta-Learning & Few-Shot Learning
Developing algorithms that can quickly adapt to new tasks with limited data, focusing on robustness and stress testing of meta-learning approaches.
Related Publications
Leveraging task variability in meta-learning
A Aimen, B Ladrecha, Sahil Sidheekh, NC Krishnan
Adaptation: Blessing or Curse for Higher Way Meta-Learning
A Aimen, Sahil Sidheekh, B Ladrecha, H Ahuja, NC Krishnan
