Introducing students to research codes: A short course on solving partial differential equations in Python

Inguva, Pavan; Bhute, Vijesh J.; Cheng, Thomas N.H.; Walker, Pierre J.

doi:10.1016/j.ece.2021.01.011

Cited by 14 publications

(3 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…or rapidly testing and optimising different designs using powerful computational resources (e.g., ASPEN (Ruiz-Ramos et al, 2017), CAD, FLUENT or COMSOL modelling, etc. (Inguva et al, 2020)).…”

Section: Nature Of Experiments In Teaching Laboratoriesmentioning

confidence: 99%

Transforming traditional teaching laboratories for effective remote delivery—A review

Bhute

Inguva

Shah

et al. 2021

Education for Chemical Engineers

Self Cite

View full text Add to dashboard Cite

Section: Nature Of Experiments In Teaching Laboratoriesmentioning

confidence: 99%

Transforming traditional teaching laboratories for effective remote delivery—A review

Bhute

Inguva

Shah

et al. 2021

Education for Chemical Engineers

Self Cite

View full text Add to dashboard Cite

“…Large language models (LLMs) based on the general framework of decoder-only autoregressive transformer models − provide powerful tools for scientific exploration. The specific class of Generative Pretrained Transformer (GPT) models has received significant attention across many fields of inquiry, suggesting remarkable possibilities, especially for generative forward and inverse tasks. ,− However, our understanding of their behavior is still in its infancy, and issues remain with respect to fact recall and potential hallucination, which requires careful validation of their predictions and a thorough exploration of implications. − There have been several recent developments that propose the use of LLM-type models, − and more generally attention-based transformer architectures, to capture the behavior of physical systems, including materials. − Other studies have suggested the broad applicability of transformer models in use cases ranging from protein folding, protein property prediction, − mechanical field predictions, , as optimizers, materials design, ,,,, and as educational tools, − among many others. The availability of large models is facilitated by recent developments such as the GPT-4 model, and also releases of open-source LLMs such as Llama-2 or Falcon. , Here, we discuss strategies for how LLMs can be improved to provide more accurate responses, especially in the context of materials analysis and design, and what mechanisms we can use to elicit more nuanced, detailed, and relevant outcomes, including interpretability.…”

Section: Introductionmentioning

confidence: 99%

Generative Retrieval-Augmented Ontologic Graph and Multiagent Strategies for Interpretive Large Language Model-Based Materials Design

Buehler

2024

ACS Eng. Au

View full text Add to dashboard Cite

Transformer neural networks show promising capabilities, in particular for uses in materials analysis, design, and manufacturing, including their capacity to work effectively with human language, symbols, code, and numerical data. Here, we explore the use of large language models (LLMs) as a tool that can support engineering analysis of materials, applied to retrieving key information about subject areas, developing research hypotheses, discovery of mechanistic relationships across disparate areas of knowledge, and writing and executing simulation codes for active knowledge generation based on physical ground truths. Moreover, when used as sets of AI agents with specific features, capabilities, and instructions, LLMs can provide powerful problem-solution strategies for applications in analysis and design problems. Our experiments focus on using a fine-tuned model, MechGPT, developed based on training data in the mechanics of materials domain. We first affirm how fine-tuning endows LLMs with a reasonable understanding of subject area knowledge. However, when queried outside the context of learned matter, LLMs can have difficulty recalling correct information and may hallucinate. We show how this can be addressed using retrievalaugmented Ontological Knowledge Graph strategies. The graph-based strategy helps us not only to discern how the model understands what concepts are important but also how they are related, which significantly improves generative performance and also naturally allows for injection of new and augmented data sources into generative AI algorithms. We find that the additional feature of relatedness provides advantages over regular retrieval augmentation approaches and not only improves LLM performance but also provides mechanistic insights for exploration of a material design process. Illustrated for a use case of relating distinct areas of knowledge, here, music and proteins, such strategies can also provide an interpretable graph structure with rich information at the node, edge, and subgraph level that provides specific insights into mechanisms and relationships. We discuss other approaches to improve generative qualities, including nonlinear sampling strategies and agent-based modeling that offer enhancements over singleshot generations, whereby LLMs are used to both generate content and assess content against an objective target. Examples provided include complex question answering, code generation, and execution in the context of automated force-field development from actively learned density functional theory (DFT) modeling and data analysis.

show abstract

“…Open-source software is readily available to the users compared to commercial software, which can be highly expensive. Moreover, open-source tools have demonstrated their scalability through their frequent use in industry and academia (Inguva et al 2021). Python is the most popular programing language in machine learning applications.…”

Section: Introductionmentioning

confidence: 99%

Modeling and optimization of a continuous electrocoagulation process using an artificial intelligence approach

2021

View full text Add to dashboard Cite

An artificial neural network (ANN) with the topology 8-94-85-2 (input – hidden layer 1 - hidden layer 2 - output) was used to model the operation of the continuous electrocoagulation (CEC) process for the removal of fluoride from water. After the ANN training, the sum of the squared errors (MSE) and the determination coefficient (R2) of the testing set model predictions were 0.0088 and 0.999, respectively, showing a good generalization and model's predictive capacity. The optimization of the process cost using the genetic algorithm (GA) showed that the optimal conditions are highly dependent on the feed concentration and the fluoride removal requirements. For a 5 L of water containing 10 mg/L of fluoride, the optimal conditions to reduce the fluoride concentration below the permissible limit (1.5 mg/L) are 88.3 mA of current intensity, a flow rate of 73.6 mL/min, and the use of a series monopolar (SM) electrode configuration, corresponding to a fluoride removal of 85% and an operating cost of 0.05 €/L.

show abstract

Introducing students to research codes: A short course on solving partial differential equations in Python

Cited by 14 publications

References 29 publications

Transforming traditional teaching laboratories for effective remote delivery—A review

Transforming traditional teaching laboratories for effective remote delivery—A review

Generative Retrieval-Augmented Ontologic Graph and Multiagent Strategies for Interpretive Large Language Model-Based Materials Design

Modeling and optimization of a continuous electrocoagulation process using an artificial intelligence approach

Contact Info

Product

Resources

About