Pawan Kumar

Prithviraj Verma, Pawan Kumar, Chandan Deshani et al.

Diffusion-weighted imaging (DWI) is used for whole-body cancer screening, but it typically requires a long acquisition time. When the scan time is reduced, the image quality often suffers, leading to increased noise in the scans. Magnitude reconstruction in DWI introduces signal-dependent Rician noise, which makes denoising more challenging for conventional convolution-based methods. To address this limitation, we propose a noise-aware attention-driven denoising framework that integrates hierarchical Swin Transformer window attention with transformer-based multi-dimensional gated refinement for DWI restoration. The model incorporates explicit noise-level conditioning and residual reconstruction to enable adaptive suppression of heteroscedastic noise across a wide range of corruption levels. Experimental evaluation on corrupted DWI scans demonstrates strong restoration performance. Our model achieves a mean PSNR of 33.69~dB and SSIM of 0.8539 across noise levels from 1\% to 15\%, while maintaining stable behavior under severe noise conditions. These results indicate that attention-guided contextual modeling combined with channel-adaptive refinement provides a robust and generalizable solution for DWI denoising.

Istasis Mishra, Arpan Dasgupta, Pratik Jawanpuria et al. · microsoft-research

Extreme multi-label (XML) classification refers to the task of supervised multi-label learning that involves a large number of labels. Hence, scalability of the classifier with increasing label dimension is an important consideration. In this paper, we develop a method called LightDXML which modifies the recently developed deep learning based XML framework by using label embeddings instead of feature embedding for negative sampling and iterating cyclically through three major phases: (1) proxy training of label embeddings (2) shortlisting of labels for negative sampling and (3) final classifier training using the negative samples. Consequently, LightDXML also removes the requirement of a re-ranker module, thereby, leading to further savings on time and memory requirements. The proposed method achieves the best of both worlds: while the training time, model size and prediction times are on par or better compared to the tree-based methods, it attains much better prediction accuracy that is on par with the deep learning based methods. Moreover, the proposed approach achieves the best tail-label prediction accuracy over most state-of-the-art XML methods on some of the large datasets\footnote{accepted in IJCNN 2023, partial funding from MAPG grant and IIIT Seed grant at IIIT, Hyderabad, India. Code: \url{https://github.com/misterpawan/LightDXML}

Andi Han, Bamdev Mishra, Pratik Jawanpuria et al. · microsoft-research

In this paper, we study min-max optimization problems on Riemannian manifolds. We introduce a Riemannian Hamiltonian function, minimization of which serves as a proxy for solving the original min-max problems. Under the Riemannian Polyak--Łojasiewicz condition on the Hamiltonian function, its minimizer corresponds to the desired min-max saddle point. We also provide cases where this condition is satisfied. For geodesic-bilinear optimization in particular, solving the proxy problem leads to the correct search direction towards global optimality, which becomes challenging with the min-max formulation. To minimize the Hamiltonian function, we propose Riemannian Hamiltonian methods (RHM) and present their convergence analyses. We extend RHM to include consensus regularization and to the stochastic setting. We illustrate the efficacy of the proposed RHM in applications such as subspace robust Wasserstein distance, robust training of neural networks, and generative adversarial networks.

Kinal Mehta, Anuj Mahajan, Pawan Kumar

Recent advances in Reinforcement Learning (RL) have led to many exciting applications. These advancements have been driven by improvements in both algorithms and engineering, which have resulted in faster training of RL agents. We present marl-jax, a multi-agent reinforcement learning software package for training and evaluating social generalization of the agents. The package is designed for training a population of agents in multi-agent environments and evaluating their ability to generalize to diverse background agents. It is built on top of DeepMind's JAX ecosystem~\cite{deepmind2020jax} and leverages the RL ecosystem developed by DeepMind. Our framework marl-jax is capable of working in cooperative and competitive, simultaneous-acting environments with multiple agents. The package offers an intuitive and user-friendly command-line interface for training a population and evaluating its generalization capabilities. In conclusion, marl-jax provides a valuable resource for researchers interested in exploring social generalization in the context of MARL. The open-source code for marl-jax is available at: \href{https://github.com/kinalmehta/marl-jax}{https://github.com/kinalmehta/marl-jax}

Arpan Dasgupta, Pawan Kumar

A large number of computational and scientific methods commonly require decomposing a sparse matrix into triangular factors as LU decomposition. A common problem faced during this decomposition is that even though the given matrix may be very sparse, the decomposition may lead to a denser triangular factors due to fill-in. A significant fill-in may lead to prohibitively larger computational costs and memory requirement during decomposition as well as during the solve phase. To this end, several heuristic sparse matrix reordering methods have been proposed to reduce fill-in before the decomposition. However, finding an optimal reordering algorithm that leads to minimal fill-in during such decomposition is known to be a NP-hard problem. A reinforcement learning based approach is proposed for this problem. The sparse matrix reordering problem is formulated as a single player game. More specifically, Monte-Carlo tree search in combination with neural network is used as a decision making algorithm to search for the best move in our game. The proposed method, alphaElimination is found to produce significantly lesser non-zeros in the LU decomposition as compared to existing state-of-the-art heuristic algorithms with little to no increase in overall running time of the algorithm. The code for the project will be publicly available here\footnote{\url{https://github.com/misterpawan/alphaEliminationPaper}}.

Ji Wei Yoon, Adithya Kumar, Pawan Kumar et al.

The combination of high-throughput experimentation techniques and machine learning (ML) has recently ushered in a new era of accelerated material discovery, enabling the identification of materials with cutting-edge properties. However, the measurement of certain physical quantities remains challenging to automate. Specifically, meticulous process control, experimentation and laborious measurements are required to achieve optimal electrical conductivity in doped polymer materials. We propose a ML approach, which relies on readily measured absorbance spectra, to accelerate the workflow associated with measuring electrical conductivity. The classification model accurately classifies samples with a conductivity > 25 to 100 S/cm, achieving a maximum of 100 % accuracy rate. For the subset of highly conductive samples, we employed a regression model to predict their conductivities, yielding an impressive test R2 value of 0.984. We tested the models with samples of the two highest conductivities (498 and 506 S/cm) and showed that they were able to correctly classify and predict the two extrapolative conductivities at satisfactory levels of errors. The proposed ML-assisted workflow results in an improvement in the efficiency of the conductivity measurements by 89 % of the maximum achievable using our experimental techniques. Furthermore, our approach addressed the common challenge of the lack of explainability in ML models by exploiting bespoke mathematical properties of the descriptors and ML model, allowing us to gain corroborated insights into the spectral influences on conductivity. Through this study, we offer an accelerated pathway for optimizing the properties of doped polymer materials while showcasing the valuable insights that can be derived from purposeful utilization of ML in experimental science.

Kinal Mehta, Anuj Mahajan, Pawan Kumar

A reliable critic is central to on-policy actor-critic learning. But it becomes challenging to learn a reliable critic in a multi-agent sparse reward scenario due to two factors: 1) The joint action space grows exponentially with the number of agents 2) This, combined with the reward sparseness and environment noise, leads to large sample requirements for accurate learning. We show that regularising the critic with spectral normalization (SN) enables it to learn more robustly, even in multi-agent on-policy sparse reward scenarios. Our experiments show that the regularised critic is quickly able to learn from the sparse rewarding experience in the complex SMAC and RWARE domains. These findings highlight the importance of regularisation in the critic for stable learning.

Arpan Dasgupta, Preeti Lamba, Ankita Kushwaha et al.

Extreme multi-label classification or XMLC, is an active area of interest in machine learning. Compared to traditional multi-label classification, here the number of labels is extremely large, hence, the name extreme multi-label classification. Using classical one-versus-all classification does not scale in this case due to large number of labels; the same is true for any other classifier. Embedding labels and features into a lower-dimensional space is a common first step in many XMLC methods. Moreover, other issues include existence of head and tail labels, where tail labels are those that occur in a relatively small number of samples. The existence of tail labels creates issues during embedding. This area has invited application of wide range of approaches ranging from bit compression motivated from compressed sensing, tree based embeddings, deep learning based latent space embedding including using attention weights, linear algebra based embeddings such as SVD, clustering, hashing, to name a few. The community has come up with a useful set of metrics to identify correctly the prediction for head or tail labels.

Prasha Srivastava, Pawan Kumar, Zia Abbas

Various studies have shown the advantages of using Machine Learning (ML) techniques for analog and digital IC design automation and optimization. Data scarcity is still an issue for electronic designs, while training highly accurate ML models. This work proposes generating and evaluating artificial data using generative adversarial networks (GANs) for circuit data to aid and improve the accuracy of ML models trained with a small training data set. The training data is obtained by various simulations in the Cadence Virtuoso, HSPICE, and Microcap design environment with TSMC 180nm and 22nm CMOS technology nodes. The artificial data is generated and tested for an appropriate set of analog and digital circuits. The experimental results show that the proposed artificial data generation significantly improves ML models and reduces the percentage error by more than 50\% of the original percentage error, which were previously trained with insufficient data. Furthermore, this research aims to contribute to the extensive application of AI/ML in the field of VLSI design and technology by relieving the training data availability-related challenges.

Neel Mishra, Pawan Kumar

In our work, we propose a novel yet simple approach to obtain an adaptive learning rate for gradient-based descent methods on classification tasks. Instead of the traditional approach of selecting adaptive learning rates via the decayed expectation of gradient-based terms, we use the angle between the current gradient and the new gradient: this new gradient is computed from the direction orthogonal to the current gradient, which further helps us in determining a better adaptive learning rate based on angle history, thereby, leading to relatively better accuracy compared to the existing state-of-the-art optimizers. On a wide variety of benchmark datasets with prominent image classification architectures such as ResNet, DenseNet, EfficientNet, and VGG, we find that our method leads to the highest accuracy in most of the datasets. Moreover, we prove that our method is convergent.

Pawan Kumar

A graph based matching is used to construct aggregation for algebraic multigrid. Effects of inexact coarse grid solve is analyzed numerically for a highly discontinuous convection diffusion coefficient matrix and problems from Florida matrix market collection. The proposed strategy is found to be more robust compared to a classical AMG approach.

Pawan Kumar

In the context of non overlapping domain decomposition methods, several algebraic approximations of the Dirichlet-to-Neumann (DtN) map are proposed in [F. X. Roux, et. al. Algebraic approximation of Dirichlet- to-Neumann maps for the equations of linear elasticity, Comput. Methods Appl. Mech. Engrg., 195, 2006, 3742-3759]. For the case of non overlapping domains, approximation to the DtN are analogous to the approximation of the Schur complements in the incomplete multilevel block factorization. In this work, several original and purely algebraic (based on graph of the matrix) domain decomposition techniques are investigated for steady state incompressible Navier-Stokes equation defined on uniform and stretched grid for low viscosity. Moreover, the methods proposed are highly parallel during both setup and application phase. Spectral and numerical analysis of the methods are also presented.

Sachin Kumar Danisetty, Santhosh Reddy Mylaram, Pawan Kumar

We propose a second order gradient based method with ADAM and RMSprop for the training of generative adversarial networks. The proposed method is fastest to obtain similar accuracy when compared to prominent second order methods. Unlike state-of-the-art recent methods, it does not require solving a linear system, or it does not require additional mixed second derivative terms. We derive the fixed point iteration corresponding to proposed method, and show that the proposed method is convergent. The proposed method produces better or comparable inception scores, and comparable quality of images compared to other recently proposed state-of-the-art second order methods. Compared to first order methods such as ADAM, it produces significantly better inception scores. The proposed method is compared and validated on popular datasets such as FFHQ, LSUN, CIFAR10, MNIST, and Fashion MNIST for image generation tasks\footnote{Accepted in IJCNN 2023}. Codes: \url{https://github.com/misterpawan/acom}

Sajal Khandelwal, Pawan Kumar, Syed Azeemuddin

Radio frequency (RF) biosensors, in particular those based on inter-digitated capacitors (IDCs), are pivotal in areas like biomedical diagnosis, remote sensing, and wireless communication. Despite their advantages of low cost and easy fabrication, their sensitivity can be hindered by design imperfections, environmental factors, and circuit noise. This paper investigates enhancing the sensitivity of IDC-based RF sensors using novel reinforcement learning based Binary Particle Swarm Optimization (RLBPSO), and it is compared to Ant Colony Optimization (ACO), and other state-of-the-art methods. By focusing on optimizing design parameters like electrode design and finger width, the proposed study found notable improvements in sensor sensitivity. The proposed RLBPSO method shows best optimized design for various frequency ranges when compared to current state-of-the-art methods.

Jérôme Kunegis, Pawan Kumar, Jun Sun et al.

We describe SynGraphy, a method for visually summarising the structure of large network datasets that works by drawing smaller graphs generated to have similar structural properties to the input graphs. Visualising complex networks is crucial to understand and make sense of networked data and the relationships it represents. Due to the large size of many networks, visualisation is extremely difficult; the simple method of drawing large networks like those of Facebook or Twitter leads to graphics that convey little or no information. While modern graph layout algorithms can scale computationally to large networks, their output tends to a common "hairball" look, which makes it difficult to even distinguish different graphs from each other. Graph sampling and graph coarsening techniques partially address these limitations but they are only able to preserve a subset of the properties of the original graphs. In this paper we take the problem of visualising large graphs from a novel perspective: we leave the original graph's nodes and edges behind, and instead summarise its properties such as the clustering coefficient and bipartivity by generating a completely new graph whose structural properties match that of the original graph. To verify the utility of this approach as compared to other graph visualisation algorithms, we perform an experimental evaluation in which we repeatedly asked experimental subjects (professionals in graph mining and related areas) to determine which of two given graphs has a given structural property and then assess which visualisation algorithm helped in identifying the correct answer. Our summarisation approach SynGraphy compares favourably to other techniques on a variety of networks.

Pawan Kumar, Hokeun Kim

Precision agriculture promises higher yields and sustainability, but adoption is slowed by the high cost of cyber-physical systems (CPS) and the lack of systematic design methods. We present a cost-aware design space exploration (DSE) framework for multimodal drone-rover platforms to integrate budget, energy, sensing, payload, computation, and communication constraints. Using integer linear programming (ILP) with SAT-based verification, our approach trades off among cost, coverage, and payload while ensuring constraint compliance and a multitude of alternatives. We conduct case studies on smaller and larger-sized farms to show that our method consistently achieves full coverage within budget while maximizing payload efficiency, outperforming state-of-the-art CPS DSE approaches.

Pawan Kumar

The discretization of Cahn-Hilliard equation with obstacle potential leads to a block 2 by 2 non-linear system, where the p1, 1q block has a non-linear and non-smooth term. Recently a globally convergent Newton Schur method was proposed for the non-linear Schur complement corresponding to this non-linear system. The solver may be seen as an inexact Uzawa method which has the falvour of an active set method in the sense that the active sets are first identified by solving a quadratic obstacle problem corresponding to the p1, 1q block of the block 2 by 2 nonlinear system, and a new decent direction is obtained after discarding the active set region. The problem becomes linear on nonactive set, and corresponds to solving a linear saddle point problem on truncated domains. For solving the quadratic obstacle problem, various optimal multigrid like methods have been proposed. In this paper solvers for the truncated saddle point problem is considered. Three preconditioners are considered, two of them have block diagonal structure, and the third one has block tridiagonal structure. One of the block diagonal preconditioners is obtained by adding certain scaling of stiffness and mass matrices, whereas, the remaining two involves Schur complement. Eigenvalue bound and condition number estimates are derived for the preconditioned untruncated problem. It is shown that the extreme eigenvalues of the preconditioned truncated system remain bounded by the extreme eigenvalues of the preconditioned untruncated system. Numerical experiments confirm the optimality of the solvers.

Pawan Kumar

The phase separation processes are typically modeled by Cahn-Hilliard equations. This equation was originally introduced to model phase separation in binary alloys, where phase stands for concentration of different components in alloy. When the binary alloy under preparation is subjected to a rapid reduction in temperature below a critical temperature, it has been experimentally observed that the concentration changes from a mixed state to a visibly distinct spatially separated two phase for binary alloy. This rapid reduction in the temperature, the so-called "deep quench limit", is modeled effectively by obstacle potential. The discretization of Cahn-Hilliard equation with obstacle potential leads to a block $2 \times 2$ {\em non-linear} system, where the $(1,1)$ block has a non-linear and non-smooth term. Recently a globally convergent Newton Schur method was proposed for the non-linear Schur complement corresponding to this non-linear system. The proposed method is similar to an inexact active set method in the sense that the active sets are first approximately identified by solving a quadratic obstacle problem corresponding to the $(1,1)$ block of the block $2 \times 2$ system, and later solving a reduced linear system by annihilating the rows and columns corresponding to identified active sets. For solving the quadratic obstacle problem, various optimal multigrid like methods have been proposed. In this paper, we study a non-standard norm that is equivalent to applying a block diagonal preconditioner to the reduced linear systems. Numerical experiments confirm the optimality of the solver and convergence independent of problem parameters on sufficiently fine mesh.

Pawan Kumar

In this paper, we study a hierarchical SSOR (HSSOR) method which could be used as a standalone method or as a smoother for a two-grid method. It is found that the method leads to faster convergence compared to more costly incomplete LU (ILU(0)) with no fill-in, the SSOR, and the Block SSOR method. Moreover, for a two-grid method, numerical experiments suggests that HSSOR can be a better replacement for SSOR smoother both having no storage requirements and have no construction costs. Using Fourier analysis, ex- pressions for the eigenvalues and the condition number of HSSOR preconditioned problem is derived for the three-dimensional isotropic model problem.

Pawan Kumar

In this paper, we use the Poincare separation theorem for estimating the eigenvalues of the fine grid. We propose a randomized version of the algorithm where several different coarse grids are constructed thus leading to more comprehensive eigenvalue estimates. The proposed algorithm is suited for modern day multicore and distributed processing.

Long Phan, Alice Gatti, Ziwen Han et al. · amazon-science, apple-ml

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 2,500 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai.

Neeraj Kollepara, Snehith Kumar Chatakonda, Pawan Kumar

In this work, we announce a comprehensive well curated and opensource dataset with millions of samples for pre-college and college level problems in mathematicsand science. A preliminary set of results using transformer architecture with character to character encoding is shown. The dataset identifies some challenging problem and invites research on better architecture search

Prasha Srivastava, Pawan Kumar, Zia Abbas

Generative AI has seen remarkable growth over the past few years, with diffusion models being state-of-the-art for image generation. This study investigates the use of diffusion models in generating artificial data generation for electronic circuits for enhancing the accuracy of subsequent machine learning models in tasks such as performance assessment, design, and testing when training data is usually known to be very limited. We utilize simulations in the HSPICE design environment with 22nm CMOS technology nodes to obtain representative real training data for our proposed diffusion model. Our results demonstrate the close resemblance of synthetic data using diffusion model to real data. We validate the quality of generated data, and demonstrate that data augmentation certainly effective in predictive analysis of VLSI design for digital circuits.

Pawan Kumar, Aditi Gupta

Modern large language models (LLMs) place extraordinary pressure on memory and compute budgets, making principled compression indispensable for both deployment and continued training. We present Hierarchical Sparse Plus Low-Rank (HSS) compression, a two-stage scheme that (i) removes the largest-magnitude weights into a sparse matrix S and (ii) applies a recursive Hierarchically Sparse Separable (HSS) low-rank factorisation to the dense residual matrix. A recursive rank-reducing strategy and a reverse Cuthill-Mckee (RCM) permutation are introduced to align high weights towards the diagonal with the block-diagonal hierarchy, maximising off-diagonal compressibility (because they are touched only once). HSS is hardware-friendly: its matrix-vector multiply reduces to one sparse and a sequence of thin-matrix multiplications and can be trained end-to-end with standard optimisers. Experiments on LLaMA-7B show that targeting only the self-attention projections (1.6 B parameters of Q, K, and V matrices out of a total 7B parameters) suffices to yield large memory savings while retaining comparable state-of-the-art perplexity scores on test samples of the WikiText dataset. For example, with a 30\% sparsity budget and an outer rank of 512, sHSS-RCM achieves a perplexity of 1.64, outperforming dense baselines and classical sparse-plus-SVD variants, while also achieving significant memory savings.

Neel Mishra, Kushagara Trivedi, Pawan Kumar

Distributed training of large neural networks is bottlenecked by full-precision gradient communication and by coordinatewise optimizers that ignore the matrix structure of weight tensors. We propose Sign-Muon, a 1-bit, matrix-aware optimizer that combines majority-vote sign aggregation from signSGD with the polar-step framework of Muon. Each worker forms a Muon-style direction by taking the polar factor of its momentum via a Newton--Schulz iteration, transmits only the entrywise signs, and aggregates by majority vote; an optional local polar step further enforces orthogonality at no extra communication cost. Under spectral-norm smoothness and bounded-variance stochastic gradients, the spectral-norm normalized sign step yields an $\mathcal{O}(1/\sqrt{T})$ nonconvex rate for an $\ell_1$-based stationarity measure. With unimodal symmetric noise, majority vote across $M$ workers cuts the stochastic term by $1/\sqrt{M}$, matching signSGD. In the $α$-$β$ model, distributed Sign-Muon needs only one integer sum-allreduce per iteration; all orthogonalization is local, giving a $32\times$ bandwidth reduction over float32 ($4\times$ for int8). Across 330 CIFAR-10/ResNet-50 configurations Sign-Muon attains the best validation accuracy (92.15\%); its 4-GPU majority-vote variant reaches 92.02\% with 37\% less training time at matched effective batch. On nanoGPT, Sign-Muon achieves lower perplexity and better anytime performance than other sign-based baselines, with favorable weak-scaling up to 16 GPUs.

Sahibpreet Singh, Pawan Kumar

This chapter explores the complexities of sports governance, taxation, dispute resolution, and the impact of digital transformation within the sports sector. This study identifies a critical research gap regarding the integration of innovative technologies to enhance governance and talent identification in sports law. The objective is to evaluate how data-driven approaches and AI can optimize recruitment processes; also ensuring compliance with existing regulations. A comprehensive analysis of current governance structures and taxation policies,(ie Income Tax Act and GST Act), reveals preliminary results indicating that reform is necessary to support sustainable growth in the sports economy. Key findings demonstrate that AI enhances player evaluation by minimizing biases and expanding access to diverse talent pools. While the Court of Arbitration for Sport provides an efficient mechanism for dispute resolution. The implications emphasize the need for regulatory reforms that align taxation policies with international best practices, promoting transparency and accountability in sports organizations. This research contributes valuable insights into the evolving dynamics of sports management, aiming to foster innovation and integrity in the industry.

Ankita Kushwaha, Kiran Ravish, Preeti Lamba et al.

Safe Reinforcement Learning (SafeRL) is the subfield of reinforcement learning that explicitly deals with safety constraints during the learning and deployment of agents. This survey provides a mathematically rigorous overview of SafeRL formulations based on Constrained Markov Decision Processes (CMDPs) and extensions to Multi-Agent Safe RL (SafeMARL). We review theoretical foundations of CMDPs, covering definitions, constrained optimization techniques, and fundamental theorems. We then summarize state-of-the-art algorithms in SafeRL for single agents, including policy gradient methods with safety guarantees and safe exploration strategies, as well as recent advances in SafeMARL for cooperative and competitive settings. Additionally, we propose five open research problems to advance the field, with three focusing on SafeMARL. Each problem is described with motivation, key challenges, and related prior work. This survey is intended as a technical guide for researchers interested in SafeRL and SafeMARL, highlighting key concepts, methods, and open future research directions.

Neel Mishra, Bamdev Mishra, Pratik Jawanpuria et al.

A novel first-order method is proposed for training generative adversarial networks (GANs). It modifies the Gauss-Newton method to approximate the min-max Hessian and uses the Sherman-Morrison inversion formula to calculate the inverse. The method corresponds to a fixed-point method that ensures necessary contraction. To evaluate its effectiveness, numerical experiments are conducted on various datasets commonly used in image generation tasks, such as MNIST, Fashion MNIST, CIFAR10, FFHQ, and LSUN. Our method is capable of generating high-fidelity images with greater diversity across multiple datasets. It also achieves the highest inception score for CIFAR10 among all compared methods, including state-of-the-art second-order methods. Additionally, its execution time is comparable to that of first-order min-max methods.

Venkata Siddharth Dhara, Pawan Kumar

In recent times, denoising diffusion probabilistic models (DPMs) have proven effective for medical image generation and denoising, and as representation learners for downstream segmentation. However, segmentation performance is limited by the need for dense pixel-wise labels, which are expensive, time-consuming, and require expert knowledge. We propose FastTextDiff, a label-efficient diffusion-based segmentation model that integrates medical text annotations to enhance semantic representations. Our approach uses ModernBERT, a transformer capable of processing long clinical notes, to tightly link textual annotations with semantic content in medical images. Trained on MIMIC-III and MIMIC-IV, ModernBERT encodes clinical knowledge that guides cross-modal attention between visual and textual features. This study validates ModernBERT as a fast, scalable alternative to Clinical BioBERT in diffusion-based segmentation pipelines and highlights the promise of multi-modal techniques for medical image analysis. By replacing Clinical BioBERT with ModernBERT, FastTextDiff benefits from FlashAttention 2, an alternating attention mechanism, and a 2-trillion-token corpus, improving both segmentation accuracy and training efficiency over traditional diffusion-based models.

Preeti Lamba, Kiran Ravish, Ankita Kushwaha et al.

Diffusion models have emerged as leading generative models for images and other modalities, but aligning their outputs with human preferences and safety constraints remains a critical challenge. This thesis proposal investigates methods to align diffusion models using reinforcement learning (RL) and reward modeling. We survey recent advances in fine-tuning text-to-image diffusion models with human feedback, including reinforcement learning from human and AI feedback, direct preference optimization, and differentiable reward approaches. We classify these methods based on the type of feedback (human, automated, binary or ranked preferences), the fine-tuning technique (policy gradient, reward-weighted likelihood, direct backpropagation, etc.), and their efficiency and safety outcomes. We compare key algorithms and frameworks, highlighting how they improve alignment with user intent or safety standards, and discuss inter-relationships such as how newer methods build on or diverge from earlier ones. Based on the survey, we identify five promising research directions for the next two years: (1) multi-objective alignment with combined rewards, (2) efficient human feedback usage and active learning, (3) robust safety alignment against adversarial inputs, (4) continual and online alignment of diffusion models, and (5) interpretable and trustworthy reward modeling for generative images. Each direction is elaborated with its problem statement, challenges, related work, and a proposed research plan. The proposal is organized as a comprehensive document with literature review, comparative tables of methods, and detailed research plans, aiming to contribute new insights and techniques for safer and value-aligned diffusion-based generative AI.

Tanmay Kumar Sinha, Jayadev Naram, Pawan Kumar

Recent approaches to the tensor completion problem have often overlooked the nonnegative structure of the data. We consider the problem of learning a nonnegative low-rank tensor, and using duality theory, we propose a novel factorization of such tensors. The factorization decouples the nonnegative constraints from the low-rank constraints. The resulting problem is an optimization problem on manifolds, and we propose a variant of Riemannian conjugate gradients to solve it. We test the proposed algorithm across various tasks such as colour image inpainting, video completion, and hyperspectral image completion. Experimental results show that the proposed method outperforms many state-of-the-art tensor completion algorithms.

Jayadev Naram, Tanmay Kumar Sinha, Pawan Kumar

We consider the problem of learning low-rank tensors from partial observations with structural constraints, and propose a novel factorization of such tensors, which leads to a simpler optimization problem. The resulting problem is an optimization problem on manifolds. We develop first-order and second-order Riemannian optimization algorithms to solve it. The duality gap for the resulting problem is derived, and we experimentally verify the correctness of the proposed algorithm. We demonstrate the algorithm on nonnegative constraints and Hankel constraints.

Pawan Kumar

As a big data application, extreme multilabel classification has emerged as an important research topic with applications in ranking and recommendation of products and items. A scalable hybrid distributed and shared memory implementation of extreme classification for large scale ranking and recommendation is proposed. In particular, the implementation is a mix of message passing using MPI across nodes and using multithreading on the nodes using OpenMP. The expression for communication latency and communication volume is derived. Parallelism using work-span model is derived for shared memory architecture. This throws light on the expected scalability of similar extreme classification methods. Experiments show that the implementation is relatively faster to train and test on some large datasets. In some cases, model size is relatively small.

Jayadev Naram, Tanmay Kumar Sinha, Pawan Kumar

We propose a novel Riemannian method for solving the Extreme multi-label classification problem that exploits the geometric structure of the sparse low-dimensional local embedding models. A constrained optimization problem is formulated as an optimization problem on matrix manifold and solved using a Riemannian optimization method. The proposed approach is tested on several real world large scale multi-label datasets and its usefulness is demonstrated through numerical experiments. The numerical experiments suggest that the proposed method is fastest to train and has least model size among the embedding-based methods. An outline of the proof of convergence for the proposed Riemannian optimization method is also stated.

Pawan Kumar, Srikanta Bedathur

Different from previous surveys in semantic parsing (Kamath and Das, 2018) and knowledge base question answering(KBQA)(Chakraborty et al., 2019; Zhu et al., 2019; Hoffner et al., 2017) we try to takes a different perspective on the study of semantic parsing. Specifically, we will focus on (a)meaning composition from syntactical structure(Partee, 1975), and (b) the ability of semantic parsers to handle lexical variation given the context of a knowledge base (KB). In the following section after an introduction of the field of semantic parsing and its uses in KBQA, we will describe meaning representation using grammar formalism CCG (Steedman, 1996). We will discuss semantic composition using formal languages in Section 2. In section 3 we will consider systems that uses formal languages e.g. $λ$-calculus (Steedman, 1996), $λ$-DCS (Liang, 2013). Section 4 and 5 consider semantic parser using structured-language for logical form. Section 6 is on different benchmark datasets ComplexQuestions (Bao et al.,2016) and GraphQuestions (Su et al., 2016) that can be used to evaluate semantic parser on their ability to answer complex questions that are highly compositional in nature.

Pawan Kumar, Dhanajit Brahma, Harish Karnick et al.

We present an attention-based ranking framework for learning to order sentences given a paragraph. Our framework is built on a bidirectional sentence encoder and a self-attention based transformer network to obtain an input order invariant representation of paragraphs. Moreover, it allows seamless training using a variety of ranking based loss functions, such as pointwise, pairwise, and listwise ranking. We apply our framework on two tasks: Sentence Ordering and Order Discrimination. Our framework outperforms various state-of-the-art methods on these tasks on a variety of evaluation metrics. We also show that it achieves better results when using pairwise and listwise ranking losses, rather than the pointwise ranking loss, which suggests that incorporating relative positions of two or more sentences in the loss function contributes to better learning.

Amartya Sanyal, Pawan Kumar, Purushottam Kar et al.

We present a class of algorithms capable of directly training deep neural networks with respect to large families of task-specific performance measures such as the F-measure and the Kullback-Leibler divergence that are structured and non-decomposable. This presents a departure from standard deep learning techniques that typically use squared or cross-entropy loss functions (that are decomposable) to train neural networks. We demonstrate that directly training with task-specific loss functions yields much faster and more stable convergence across problems and datasets. Our proposed algorithms and implementations have several novel features including (i) convergence to first order stationary points despite optimizing complex objective functions; (ii) use of fewer training samples to achieve a desired level of convergence, (iii) a substantial reduction in training time, and (iv) a seamless integration of our implementation into existing symbolic gradient frameworks. We implement our techniques on a variety of deep architectures including multi-layer perceptrons and recurrent neural networks and show that on a variety of benchmark and real data sets, our algorithms outperform traditional approaches to training deep networks, as well as some recent approaches to task-specific training of neural networks.