Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Engage Grants Program

Université Laval

$25,000.00

Machine learning for mine exploration target discovery

FY 2016-17

Summary
SGS Geostat has built a strong expertise in geological modelling software and has currently a tiny expertsystem for computing a prospectivity score (for a given metal such as gold) of set of regions which are oftenexpressed as a 3D block model of 50 m by 50 m by 50 m. The expert system's prospectivity score for eachblock is based on the values of 14 variables which are sometimes obtained from drilling measurements (whenavailable) but, most often, are obtained by interpolation or extrapolation of existing data from geological andgeophysical models.The research project consists in investigating the efficiency of state-of-the-art machine learning algorithms forconstructing predictors of the prospectivity score of these blocks in comparison to what can be obtained fromthe current expert system predictions. Since the expert system consists of rules and variables that can be usedas features for machine learning algorithms and that many more informative features can be extracted from thegeological model, we plan to use Tikhonov regularization approaches such as ridge regression and logisticregression since they can be used with a very large set of features and adaptive weights without suffering fromthe curse of dimensionality since the prediction guarantees do not depend on the number of features used (ie,the dimensionality) but only on the Euclidean norms of the predictor's weight vector and of the feature vector.Moreover, because of the strong convexity of the L2 regularizer used in these approaches, overfitting is easilycontrolled by a hyperparameter. If the quality of the predictions are not satisfactory, we also plan to train morecomplex models (on the same set of features) such as neural networks. Finally, to investigate if we can builtsparser and more easily-interpretable models, we plan to use boosting methods and regression tree approachessuch as CART.
Program
Engage Grants Program
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
500218-2016
Name nom
Marchand, Mario
Department département
Informatique et de génie logiciel
Organization id
51
Institution établissement
Université Laval
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2016
Competition year année de concours
2016
Award amount
25000
Program id
EGP
Program name en
Engage Grants Program
Program name fr
Programme de subventions d'engagement partenariat
Group en
Research partnerships programs
Group fr
Programmes de partenariat
Committee code
1552
Committee name en
Quebec Internal Decision Committee
Committee name fr
Comité de décision interne pour le Québec
Area of application code
202
Area of application group en
Natural resources (economic aspects)
Area of application group fr
Ressources naturelles (aspects économiques)
Area of application en
Mineral resources ( prospecting, exploration, mining, extraction, processing)
Area of application fr
Ressources minérales (prospection, exploration, mines, extraction et préparation)
Research subject code
2805
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Learning and inference theories
Research subject fr
Théories de l'apprentissage et l'inférence
Partie
1
Nb partie
1
Application title
Machine learning for mine exploration target discovery
Keyword
NSERC
Application summary
SGS Geostat has built a strong expertise in geological modelling software and has currently a tiny expertsystem for computing a prospectivity score (for a given metal such as gold) of set of regions which are oftenexpressed as a 3D block model of 50 m by 50 m by 50 m. The expert system's prospectivity score for eachblock is based on the values of 14 variables which are sometimes obtained from drilling measurements (whenavailable) but, most often, are obtained by interpolation or extrapolation of existing data from geological andgeophysical models.The research project consists in investigating the efficiency of state-of-the-art machine learning algorithms forconstructing predictors of the prospectivity score of these blocks in comparison to what can be obtained fromthe current expert system predictions. Since the expert system consists of rules and variables that can be usedas features for machine learning algorithms and that many more informative features can be extracted from thegeological model, we plan to use Tikhonov regularization approaches such as ridge regression and logisticregression since they can be used with a very large set of features and adaptive weights without suffering fromthe curse of dimensionality since the prediction guarantees do not depend on the number of features used (ie,the dimensionality) but only on the Euclidean norms of the predictor's weight vector and of the feature vector.Moreover, because of the strong convexity of the L2 regularizer used in these approaches, overfitting is easilycontrolled by a hyperparameter. If the quality of the predictions are not satisfactory, we also plan to train morecomplex models (on the same set of features) such as neural networks. Finally, to investigate if we can builtsparser and more easily-interpretable models, we plan to use boosting methods and regression tree approachessuch as CART.