arXiv subjects

Ajitabh Kumar

Publications and source records attributed to Ajitabh Kumar.

6 recordsLinked to original sources

arXiv2026

Honest Small Decision Models for Finance: Scale, Specialization Collapse, and Mixed-Corpus Remediation

Financial workflows need fast, calibrated, auditable decisions, not just generated text. We train System-1 decision models, small language models fine-tuned to answer typed questions (choice/score/noul) with a probability in a single forward pass, across two model families (K2-Horizon and Qwen3.5/3.8) and nine rungs from 0.8B to 32B parameters, using one frozen LoRA recipe on a public decision benchmark and two new finance corpora: a programmatic-gold policy corpus (rule-reading with exact labels) and a reality-labeled outcome corpus (19,640 realized forward-return questions in walk-forward splits, plus a sealed 2026 test of 1,360 questions). Narrow fine-tuning costs general decision competence 9-45 percentage points, a data-mixture effect that shrinks with model scale and is associated with narrow-gradient volume. On reality-labeled questions, every trained rung from 0.8B to 32B lands at the horizon-conditioned prior baseline (accuracy 0.37-0.38, no detectable conditional edge under our controls); the models respond with flat, well-calibrated probabilities that track per-horizon base rates: naturally honest uncertainty, with a mid-scale calibration sweet spot (ECE 0.026 at 2B, degrading to 0.043-0.044 at 9-32B under prior drift). A mixed-corpus remediation carries all competencies in one adapter at modest cost: it matches or slightly exceeds the generalist specialist on the public benchmark while holding the policy ceiling and near-specialist outcome calibration, yielding a deployable artifact with per-domain temperature regimes and a band-dispatch escalation analysis: sharp, dispatchable bands where rules determine answers; flat, review-routing bands where reality does not.

cs.LG↗

arXiv2023

Precipitation Nowcasting With Spatial And Temporal Transfer Learning Using Swin-UNETR

Climate change has led to an increase in frequency of extreme weather events. Early warning systems can prevent disasters and loss of life. Managing such events remain a challenge for both public and private institutions. Precipitation nowcasting can help relevant institutions to better prepare for such events. Numerical weather prediction (NWP) has traditionally been used to make physics based forecasting, and recently deep learning based approaches have been used to reduce turn-around time for nowcasting. In this work, recently proposed Swin-UNETR (Swin UNEt TRansformer) is used for precipitation nowcasting for ten different regions of Europe. Swin-UNETR utilizes a U-shaped network within which a swin transformer-based encoder extracts multi-scale features from multiple input channels of satellite image, while CNN-based decoder makes the prediction. Trained model is capable of nowcasting not only for the regions for which data is available, but can also be used for new regions for which data is not available.

physics.ao-ph↗

arXiv2023

Skilful Precipitation Nowcasting Using NowcastNet

Designing early warning system for precipitation requires accurate short-term forecasting system. Climate change has led to an increase in frequency of extreme weather events, and hence such systems can prevent disasters and loss of life. Managing such events remain a challenge for both public and private institutions. Precipitation nowcasting can help relevant institutions to better prepare for such events as they impact agriculture, transport, public health and safety, etc. Physics-based numerical weather prediction (NWP) is unable to perform well for nowcasting because of large computational turn-around time. Deep-learning based models on the other hand are able to give predictions within seconds. We use recently proposed NowcastNet, a physics-conditioned deep generative network, to forecast precipitation for different regions of Europe using satellite images. Both spatial and temporal transfer learning is done by forecasting for the unseen regions and year. Model makes realistic predictions and is able to outperform baseline for such a prediction task.

physics.ao-ph↗

arXiv2022

Surrogate Model For Field Optimization Using Beta-VAE Based Regression

Oilfield development related decisions are made using reservoir simulation-based optimization study in which different production scenarios and well controls are compared. Such simulations are computationally expensive and so surrogate models are used to accelerate studies. Deep learning has been used in past to generate surrogates, but such models often fail to quantify prediction uncertainty and are not interpretable. In this work, beta-VAE based regression is proposed to generate simulation surrogates for use in optimization workflow. beta-VAE enables interpretable, factorized representation of decision variables in latent space, which is then further used for regression. Probabilistic dense layers are used to quantify prediction uncertainty and enable approximate Bayesian inference. Surrogate model developed using beta-VAE based regression finds interpretable and relevant latent representation. A reasonable value of beta ensures a good balance between factor disentanglement and reconstruction. Probabilistic dense layer helps in quantifying predicted uncertainty for objective function, which is then used to decide whether full-physics simulation is required for a case.

cs.LG↗

arXiv2021

Multi-stage, multi-swarm PSO for joint optimization of well placement and control

Evolutionary optimization algorithms, including particle swarm optimization (PSO), have been successfully applied in oil industry for production planning and control. Such optimization studies are quite challenging due to large number of decision variables, production scenarios, and subsurface uncertainties. In this work, a multi-stage, multi-swarm PSO (MS2PSO) is proposed to fix certain issues with canonical PSO algorithm such as premature convergence, excessive influence of global best solution, and oscillation. Multiple experiments are conducted using Olympus benchmark to compare the efficacy of algorithms. Canonical PSO hyperparameters are first tuned to prioritize exploration in early phase and exploitation in late phase. Next, a two-stage multi-swarm PSO (2SPSO) is used where multiple-swarms of the first stage collapse into a single swarm in the second stage. Finally, MS2PSO with multiple stages and multiple swarms is used in which swarms recursively collapse after each stage. Multiple swarm strategy ensures that diversity is retained within the population and multiple modes are explored. Staging ensures that local optima found during initial stage does not lead to premature convergence. Optimization test case comprises of 90 control variables and a twenty year period of flow simulation. It is observed that different algorithm designs have their own benefits and drawbacks. Multiple swarms and stages help algorithm to move away from local optima, but at the same time they may also necessitate larger number of iterations for convergence. Both 2SPSO and MS2PSO are found to be helpful for problems with high dimensions and multiple modes where greater degree of exploration is desired.

cs.NE↗

arXiv2021

Hybrid Evolutionary Optimization Approach for Oilfield Well Control Optimization

Oilfield production optimization is challenging due to subsurface model complexity and associated non-linearity, large number of control parameters, large number of production scenarios, and subsurface uncertainties. Optimization involves time-consuming reservoir simulation studies to compare different production scenarios and settings. This paper presents efficacy of two hybrid evolutionary optimization approaches for well control optimization of a waterflooding operation, and demonstrates their application using Olympus benchmark. A simpler, weighted sum of cumulative fluid (WCF) is used as objective function first, which is then replaced by net present value (NPV) of discounted cash-flow for comparison. Two popular evolutionary optimization algorithms, genetic algorithm (GA) and particle swarm optimization (PSO), are first used in standalone mode to solve well control optimization problem. Next, both GA and PSO methods are used with another popular optimization algorithm, covariance matrix adaptation-evolution strategy (CMA-ES), in hybrid mode. Hybrid optimization run is made by transferring the resulting population from one algorithm to the next as its starting population for further improvement. Approximately four thousand simulation runs are needed for standalone GA and PSO methods to converge, while six thousand runs are needed in case of two hybrid optimization modes (GA-CMA-ES and PSO-CMA-ES). To reduce turn-around time, commercial cloud computing is used and simulation workload is distributed using parallel programming. GA and PSO algorithms have a good balance between exploratory and exploitative properties, thus are able identify regions of interest. CMA-ES algorithm is able to further refine the solution using its excellent exploitative properties. Thus, GA or PSO with CMA-ES in hybrid mode yields better optimization result as compared to standalone GA or PSO algorithms.

cs.NE↗