Meet the Speaker: Amina Lamghari

 

Meet the speaker:
Assistant Professor Amina Lamghari

Ahead of her presentation at AMSI Optimise 2019, invited speaker Assistant Professor Amina Lamghari from the University of Quebec shared with us her exciting research which has opened doors for new methodologies capitalising on the synergies between artificial intelligence and optimization techniques.

Can you tell us about your work? What drives your interest in this field?

My research interests lie in modeling and solving real-life, large-scale optimization problems, with a particular interest in developing efficient algorithms for solving difficult combinatorial optimization problems that engineers and operations managers are continually confronted with. This concern has stimulated my interest in meta-heuristics, which can be seen as generic heuristic solution approaches designed to control and guide specific problem-oriented heuristics. In the last two decades, these practical algorithmic approaches, whose emergence is considered as one of the most notable recent achievements in operations research, have been the most popular techniques for solving complex problems, particularly those of combinatorial nature, such as those arising in mining.

Carrying this interest forward, recently, I have been working on hyper-heuristics, which are emergent search methodologies integrating machine learning and optimization techniques, that seek to automate the process of selecting and combining simpler heuristics or of generating new heuristics from components of existing heuristics in order to solve hard computational search problems. Like meta-heuristics, hyper-heuristics can tackle large-scale complex optimization problems and find near-optimal solutions in a reasonable amount of time, but their advantage over meta-heuristics is that they are self-managed. Not only can they tune algorithm parameters, but they can also select the solution approach that is best suited for the given input at each decision point, making them more general than the other methodologies.

What are the most interesting “big questions” or challenges facing researchers in your area?

Geological uncertainty is an inherent aspect of mine production scheduling. Explicitly accounting for it when devising a production schedule can reduce risk, enabling mining companies to meet production targets and make the best possible return on investment. However, accounting for uncertainty adds more complexity to an optimization problem that is already complex due to its combinatorial nature and large scale. Devising algorithms that can meet the challenges of scale, complexity and uncertainty and that are capable of producing high-quality solutions to complex problems within short computing times is one of the most interesting “big questions” facing researchers in my area.

What are some key industry applications of your work?

Many of the algorithms I have developed during my time at the COSMO Stochastic Mine Planning Laboratory were incorporated in the COSMO Suite software, a platform designed to aid in the dissemination of new stochastic mine planning optimization algorithms, techniques, and models to the mining industry. This software provides a graphical user interface to create models for mineral value chains, optimize them, and perform risk analyses.

What do you consider your biggest achievement to date?

Mineral value chains involve mining, processing, stockpiling, and transportation activities. Their optimization is typically partitioned into two stages considered sequentially: the first one focuses on block extraction, while the second one focuses on flow optimization. An integrated optimization of these two stages can both increase the net present value of the mining project and lead to more robust and coordinated schedules. However, it entails solving a larger and more complex combinatorial optimization problem. I tackled this complex problem with a novel hyper­-heuristic method that relies on a high-­level strategy combining elements from reinforcement learning and tabu search to guide and automate the selection of simple perturbative low­-level heuristics. Not only can the developed method efficiently solve large instances of practical interest, but it is also self­-managed and does not require problem­ specific knowledge, making it a more generally applicable methodology that provides a framework for automating the design of solution methods. This research has also opened doors for new methodologies capitalizing on the synergies between artificial intelligence and optimization techniques. The development of hybrid methods integrating components from exact algorithms (relaxation and decomposition), machine learning techniques (reinforcement learning and artificial neural networks), and heuristics (local improvement and randomized search) is a relatively recent but ongoing focus of my research and shows great potential moving forward.

After receiving a BSc and a Master’s in Applied Mathematics, Amina Lamghari obtained a PhD in Operations Research from the University of Montreal, Canada, after which she worked as a post-doctoral fellow and later as a research associate at the COSMO Stochastic Mine Planning Laboratory at McGill University. Amina is currently an assistant professor in the Management School at the University of Quebec. Her research interests are centered on various techniques and algorithms —(meta)heuristics, hyper-heuristics, and matheuristics — for optimization and their integration and application to solve complex scheduling and planning problems in an efficient manner. All of her work to date has had direct real­-world applications. During the last nine years, applications in mine planning have been the main focus of her research, undertaken in response to the needs of the mining industry for more efficient and robust production schedules and operations plans. Many of the algorithms she developed were incorporated in the COSMO Suite software, a platform designed to aid in the dissemination of new stochastic mine planning optimization algorithms, techniques, and models to the mining industry.

Meet the Speaker: Amina Lamghari

Professor Amina Lamghari

University of Quebec

Amina will present on ‘simultaneous stochastic optimization of mining complexes / mineral value chains‘ at AMSI Optimise 2019, which will be held at the Hyatt Regency Perth from 17-21 June

Recent Posts