FES Funded ProjectsOutputs
| Title |
Category |
Date |
Authors |
| Disentangling Structural Confusion through Machine Learning: Structure Prediction and Polymorphism of Equiatomic Ternary Phases ABCThis paper illustrates that apparently "ambiguous" predictions (with probabilities in an intermediate range not close to zero or one) made by machine-learning methods can be interpreted in a meaningful way by relating them to the likely occurrence of structural polymorphism, as illustrated here for ternary phases ABC. The work is an important breakthrough in machine-learning methods for structural predictions.T12-P01 University of Alberta | Publication | 2017-11-13 | Oliynyk, A., Adutwum, L., Brent W Rudyk, Harshil Pisavadia, "Sogol Lotfi ", "Viktor Hlukhyy ", James J Harynuk, Mar, A., "Jakoah Brgoch " | | How To Optimize Materials and Devices via Design of Experiments and Machine Learning: Demonstration Using Organic PhotovoltaicsThis widely publicized paper, in the form of a “Perspective,” was the result of an ongoing collaboration between the Buriak and Mar group, illustrating how machine learning is applied to optimize several experimental conditions so that efficiencies of photovoltaic devices can be rapidly improved. [To date, this paper has been viewed over 7100 times in just over a year.]T12-P01, T12-P04 University of Alberta | Publication | 2018-07-01 | | | Solving the Colouring Problem in Half-Heusler Structures: Machine-Learning Predictions and Experimental ValidationThis paper tackles a significant challenge in inorganic structural chemistry known as the “colouring problem,” namely how atoms are distribution over different sites in a crystal structure. For the first time, we successfully apply a machine-learning approach to predict the site distributions in the very large family several hundred compounds) of half-Heusler compounds, which find many applications as thermoelectric and spintronic materials. Unique to our approach, we validate these predictions with experimental measurements.T12-P01, T12-Z01 University of Alberta | Publication | 2019-06-25 | | | How large is an atom?Contributed talk. We employed a machine-learning approach to gain chemical insight on the factors that affect atomic size, with implications on how this revises some fundamental concepts in crystal chemistry such as radius ratio rules. This model also accurately predicts interatomic distances in hypothetical compounds, which can be valuable information when investigating experimentally challenging systems containing expensive or radioactive elements.T12-P01 University of Alberta | Activity | 2018-05-28 | | | Discovery of Noncentrosymmetric Ternary Compounds from Elemental Composition: A Machine-Learning ApproachContributed talk. The absence of an inversion centre in crystalline solids is an important prerequisite for many useful electrical and optical properties in materials applications such as piezoelectric and nonlinear optical devices. Here, we apply machine-learning approaches to discover new noncentrosymmetric compounds based solely on their composition.T12-P01 University of Alberta | Activity | 2018-05-28 | | | Classification of Half-Heusler Compounds through Machine Learning ApproachesContributed talk. Half-Heusler compounds have many applications as thermoelectric materials, spintronic materials, superconductors, and topological insulators. We have applied machine-learning approaches to classify, verify, and predict half-Heusler compounds.T12-P01, T12-Z01 University of Alberta | Activity | 2018-05-28 | | | Classification of Half-Heusler Compounds through Machine-Learning ApproachesContributed poster.T12-P01, T12-Z01 University of Alberta | Activity | 2018-07-25 | | | Introduction to Machine Learning: A Practical WorkshopThis full-day workshop to introduce machine learning for materials discovery and analysis was organized and offered by several members of FES.
This workshop provided an overview of machine learning applied by UofA researchers at
Engineering and Chemistry departments. The introduction to the practical application included
data processing and preparation aspects, as well as creating and running machine-learning models
on free demo version software. Participants had an opportunity to learn step-by-step how to
handle the data, use the models, and interpret the results.
This workshop was offered twice, first on August 13, 2018 and second on August 16, 2018.T12-P01 University of Alberta | Activity | 2018-08-13 | Oliynyk, A., Adutwum, L., Mar, A., "Ajay Ganesh ", "Anjana Puliyanda ", "Kaushik Sivaramakrishnan ", Gokul Sai Subraveti, Kasturi Nagesh Pai, Prasad, V. | | Effectively Exploring Parameter Space: Design of Experiments and Machine Learning-assisted Organic Solar Cell Efficiency OptimizationInvited talk at the annual American Physical Society meeting, Boston, MA
https://meetings.aps.org/Meeting/MAR19/Session/B55.7T12-P04 University of Alberta | Activity | 2019-03-04 | | | Invited talk: ACS national meeting, Boston Invited seminar at NASA symposium on space exploration. Seminar was about solar energy optimization. Talk by Buriak.T12-P04 University of Alberta | Activity | 2019-04-18 | | | Discovery of ternary noncentrosymmetric compounds: A machine-learning approach with experimental proofContributed oral presentationT12-P01 University of Alberta | Activity | 2019-06-07 | | | Accelerating the Discovery of Materials with Machine Learning: Potential Roadblocks and How to Overcome ThemContributed oral presentationT12-P01 University of Alberta | Activity | 2019-06-07 | | | Design of Experiments and Machine Learning-Assisted Organic Solar Cell Efficiency OptimizationOrganic solar cells (OSCs) represent a cost-effective way to transform solar energy into electricity due to their potential for low-cost and high-throughput roll-to-roll production.[1] Improving OSC efficiency and stability are two of the most important tasks on the way toward commercialization. While much effort has been focused on developing new materials, there is enormous room with respect to optimization of the processing of OSCs to achieve optimal performance of a particular photoactive material. Morphology of the bulk heterojunction, the most important layer within an OSC, results from nanoscale phase segregation that depends extensively upon processing. However, the optimization process for an OSC is tedious, time-consuming and expensive involving many parameters. Conventional optimization uses unimodal approach where one parameter is optimized while all others are kept constant. This can take months or years. Moreover, there is an associated risk of missing the optimal results because all experimental combinations for all parameters cannot be performed. Herein, we report an approach that uses Design of Experiments (DOE) along with machine learning statistical data analysis to effectively and efficiently optimize solar cell efficiency. Machine learning algorithms are trained to find patterns in datasets which could greatly assist data analysis and parameter importance evaluation, hence predict the results for future experiments.[2] DOE methods allow experimentalists to explore a larger parameter space with fewer experimental trials while obtaining valid and objective conclusions. We show that, using the principles of DOE to plan experiments combined with using machine learning assist in the prediction of optimized solar cells. Specific examples of concrete improvement of the power conversion efficiency of OSCs will be described. T12-P04 University of Alberta | Activity | 2018-05-25 | Bing Cao, Adutwum, L., Brian Olsen, Oliynyk, A., Erik Luber, Tate Hauger, Mar, A., Jillian Mary Buriak | | Solving the Colouring Problem in Half- Heusler Structures: Machine-Learning Predictions and Experimental ValidationPoster presentation at the North American Solid State Chemistry Conference.T12-P01 University of Alberta | Activity | 2019-07-31 | | | Predicting noncentrosymmetric quaternary tellurides using machine learningT12-P01 University of Alberta, Manhattan College, University of Ghana | Activity | 2021-07-28 | |
| |
