Taufique Atique

Very interesting and unheard of! This is the first course of Scientific Machine Learning in Oil and Gas that doesn’t call it “AI” or “Artificial Intelligence”. Match point for the scientific side! 💯 SC-15: Introduction to Scientific Machine Learning for Geophysicists Monday, 25 August | 8:00 a.m.–5:00 p.m. | George R. Brown Convention Center • Introduction to PINNs • PINNs for solving forward and inverse problems in Geophysics • Introduction to Neural Operators • Solution of forward and inverse problems using neural operators • Physics-informed Neural Operators • Emerging trends in Scientific Machine Learning #SciML #petroleumEngineering #SPE #oilandgas #PINN #PILM #PhysicsOfData #notAI

14

1 Comment

The ML review system is so broken. - Researchers write more papers than anyone cares to read - Reviewers have no interest in reviewing. So, if they don't review, the conference has to shame them to their coauthors. They choose to use LLMs to review instead of reading themselves. - Authors try to fool these unethical reviewers by prompt injection, to try to elicit only positive reviews - And why write all these papers if no one actually reads them? Because researchers are perceived and evaluated by the number of papers (people lack the time and/or expertise to read and evaluate them themselves) The whole thing is dystopic.

4,260

165 Comments

Happy to share that two of my first-authored papers have been accepted at #AISTATS 2026 (https://lnkd.in/gzFNvPZ2). Both share a non-parametric theme, but tackle very different problems. _ 1) Meta Sparse Principal Component Analysis (with, Jean Honorio) Paper: https://lnkd.in/gEqJFvt9 Modern data problems often come in many related tasks—think multiple hospitals, cohorts, devices, or time periods—where each task has very limited samples, but the overall system shares a hidden structure. A core challenge is: how do we reliably learn that shared structure when each individual task is underpowered, which is what we addressed in this paper. What's new? • A meta-learning framework for sparse PCA that focuses on recovering the important variables/features shared across tasks • Clear conditions for when cross-task transfer works and why • Strong guarantees for both multi-task learning and fast adaptation to a NEW task Why it matters: This helps push sparse PCA toward realistic, messy deployments where data are naturally split across many small groups. It’s especially relevant when: • privacy, cost, or logistics limit per-site sample sizes (e.g., healthcare, genomics), • each task is high-dimensional, and • we still want interpretable, feature-sparse structure rather than black-box representations. _ 2) Nonparametric Multi Change Point Detection for Markov Chains via Adaptive Clustering (with Jiaqi Lei, and Sanjay Mehrotra) Many time-ordered datasets aren’t independent: what happens next depends on what happened before (user behavior, markets, sensor streams, queues, clinical trajectories, etc.). That dependence makes change-point detection much harder—and it’s what we study in this paper. It is not on arxiv, but will soon be made available on PMLR. What’s new? • A practical framework based on adaptive clustering + penalization to detect multiple changes in Markovian sequences • A key technical ingredient: a concentration tool designed for regenerative structure in Markov chains, enabling rigorous finite-sample control • Guarantees that support recovering both the number and the location of change-points. • A computational angle: we also discuss implementations that connect naturally to optimization-based formulations used in practice. Why it matters? Change points in dependent sequences show up everywhere: • monitoring systems and sensors (fault detection) • finance (regime shifts) • healthcare (patient trajectory shifts), etc. My unending gratitude to Purdue University and especially to IEMS at Northwestern University, for providing me the support, resources, and freedom to chase questions I love. If you are visiting Tangier, Morocco, at that time leave a comment and I would love to connect. OD3E #AISTATS2026 #MachineLearning #Statistics #MetaLearning #SparsePCA #HighDimensionalStatistics #RepresentationLearning #ChangePointDetection #MarkovChains #Nonparametric #Optimization #DataScience

82

3 Comments

🎉 We are thrilled to announce the publication of our paper in Energies! 📄 Title: "Fault Detection and Fault-Tolerant Control of Permanent Magnet Linear Motors Using an Emotional Learning-Based Neural Network and a Linear Extended State Observer." This article introduces a robust framework that utilizes an innovative data-driven framework to decrease energy consumption while facilitating real-time Fault-Tolerant Control (FTC) and Active Fault Detection (AFD). The proposed framework represents a significant advancement toward smarter and more efficient motor control systems. Additionally, we demonstrate Lyapunov stability analysis of the framework, which ensures that the system remains stable under various operating conditions and disturbances. I am deeply grateful to Dr. Amirmehdi Yazdani and Dr. Hamed Habibi for their valuable efforts, support, and collaboration throughout this journey. This achievement would not have been possible without their guidance and dedication. 🙏 The paper is available through the following link. Please let me know if you would like to read it! https://lnkd.in/eEnvDAus #FaultDetection #FaultTolerantControl #LinearMotors #MachineLearning #EnergyEfficiency #ControlSystems #OpenAccess #Research #Energies

21

10 Comments

"Why Do Transformers Fail to Forecast Time Series In-Context?" I read the paper by Yufa Zhou, Yixiao Wang, Surbhi Goel, and Anru R. Zhang, "Why Do Transformers Fail to Forecast Time Series In-Context?" https://lnkd.in/dAeMmBmp I'm pleased to have gained theoretical justification for what I observed empirically when comparing the results of classical seismic inversions with those based on ML algorithms. From my experience, I know that the results of ML-based inversions tend to resemble those of classical seismic inversions, but their outputs are often contaminated with random noise and appear blurred. Therefore, I recommend this article not only to those like me who work in exploration seismics but also to anyone involved in time series forecasting in general. I also asked ChatGPT to summarise this article, and it provided me with the following image. Judging by the errors, the system itself generated the image, so I don't think this image infringes anyone's copyright. 😊

3

1 Comment

Agent Engineering Isn’t Prompt Engineering — It’s a New Discipline I recently read an article on “Agent Engineering as a New Discipline,” and it resonated deeply with what many of us are experiencing in production. If you’ve ever built an agent, you know this truth: 👉 The gap between “it works on my machine” and “it works in production” is massive. Agents aren’t traditional software. They’re non-deterministic, adaptive systems where: Every user input is an edge case “Working” isn’t binary 99.9% uptime can still mean a broken experience That’s why agent engineering isn’t about better prompts. It’s about making probabilistic systems reliable, observable, and trustworthy at scale. A few hard-earned lessons from the field: 🔹 Shipping is not the finish line — it’s the learning loop Production is where agents reveal their true behavior. Real reliability comes from instrumenting decisions, tool calls, and context — then iterating fast. 🔹 You can’t debug agents the old way When reasoning lives inside the model, logs aren’t enough. You need execution traces, decision lineage, and behavioral evaluation — not just pass/fail tests. 🔹 Agent engineering is a team sport It sits at the intersection of: Product (intent, JTBD, eval definitions) Engineering (tools, runtimes, human-in-the-loop) Data science (evals, drift, error patterns) But in regulated, high-impact domains, there’s a fourth pillar we can’t ignore: 👉 Governance, trust, and accountability As agents take on workflows that once required human judgment, the real question shifts from: “Can we build agents?” to “Can we trust them — and explain their decisions?” My strong belief: Agent engineering will become table stakes. Agent governance will be the differentiator. The teams that win won’t just ship agents faster — they’ll be the ones who can observe, evaluate, refine, and stand behind agent behavior in production. Curious how others are thinking about agent reliability, evals, and trust as agents move from demos to core systems.

7

Excited to share that Mahmudur Rahman and I got our paper "ColdNet: Neural Causal Inference Under Extreme Imbalance and Sparsity" accepted at the WSDM Workshop on Benchmarking Causal Models (CausalBench'26). This paper introduces a novel solution for estimating individual treatment effects under three co-occurring challenges - extreme class imbalance (0.39% treatment rate), 99%+ cold start cases(no history), and 97%+ sparse outcomes, achieving strong results in production at billion-scale. I had the pleasure of attending the conference in Boise, Idaho this past February and presenting this work in person. Big thanks to CausalBench'26 Program Chairs and organizers for hosting and seamless communication throughout each process, as well as Amazon Science for supporting this piece of work.

54

1 Comment

[2501, arXiv] Retrieval-Augmented Generation with Graphs (GraphRAG), Haoyu Han et al. - Site: https://lnkd.in/gJq-bgGh - Abstract: Retrieval-augmented generation (RAG) is a powerful technique that enhances downstream task execution by retrieving additional information, such as knowledge, skills, and tools from external sources. Graph, by its intrinsic "nodes connected by edges" nature, encodes massive heterogeneous and relational information, making it a golden resource for RAG in tremendous real-world applications. As a result, we have recently witnessed increasing attention on equipping RAG with Graph, i.e., GraphRAG. However, unlike conventional RAG, where the retriever, generator, and external data sources can be uniformly designed in the neural-embedding space, the uniqueness of graph-structured data, such as diverse-formatted and domain-specific relational knowledge, poses unique and significant challenges when designing GraphRAG for different domains. Given the broad applicability, the associated design challenges, and the recent surge in GraphRAG, a systematic and up-to-date survey of its key concepts and techniques is urgently desired. Following this motivation, we present a comprehensive and up-to-date survey on GraphRAG. Our survey first proposes a holistic GraphRAG framework by defining its key components, including query processor, retriever, organizer, generator, and data source. Furthermore, recognizing that graphs in different domains exhibit distinct relational patterns and require dedicated designs, we review GraphRAG techniques uniquely tailored to each domain. Finally, we discuss research challenges and brainstorm directions to inspire cross-disciplinary opportunities.

8

Theoretical computer science has a culture of rigor that’s hard to overstate: a single inconsistent variable, a missing quantifier, or an inequality applied one step too far can quietly derail months of work. That’s why I found this Google Research write-up so interesting: they piloted a Gemini-based “pre-submission feedback” tool for STOC 2026 papers, essentially an automated, fast-turnaround collaborator designed to pressure-test proofs and presentation before human peer review. A few takeaways that stood out to me: - When feedback arrives in a day or two (instead of after a long internal pass or a conference deadline scramble), authors can actually act on it while it still matters. - Even top researchers get snagged by “embarrassingly simple bugs.” A tool that reliably flags inconsistent notation, missing assumptions, or shaky steps is like a spellcheck, except for lemmas. - The format they used (summary, potential mistakes/improvements, minor corrections) mirrors how strong advisors and reviewers respond: first orient, then attack the core logic, then polish. - The human remains in the loop by design. Participants reportedly filtered occasional hallucinations and extracted the useful signal. That’s a healthy framing: not “AI replaces peer review,” but “AI increases the chance your draft reaches peer review in its best form.” The part I keep coming back to is the educational angle: immediate, rigorous feedback on clarity and proof structure could be a major accelerant for training graduate students (and honestly, anyone learning to write math well). If tools like this keep improving, I can imagine a near future where “proof hygiene” becomes more like unit testing: something you run early, often, and automatically, so humans spend more time on ideas and less time chasing typos that mutate into theorems. #TheoryOfComputing #Algorithms #Mathematics #ResearchWorkflow #PeerReview #GenerativeAI #AIInEducation #AcademicWriting #STOC #ComputerScience #TheoreticalComputerScience #Engineering #DataScience #Theorem #Proofs

6

1 Comment

Guest writer on my substance, Kyle Butts, discusses two papers on causal inference with spatial data for those work with data like that. https://lnkd.in/e9jqdZsT

32

1 Comment