Subject to: Tobias Achterberg
Dr. Achterberg studied mathematics and computer science at the Technical University of Berlin and the Zuse Institute Berlin. He finished his PhD in mathematics under supervision of Prof. Martin Grötschel in 2007. Dr. Achterberg is the author of SCIP, currently the best academic MIP solver. In addition to numerous publications in scientific journals he has also received several awards for his dissertation and for SCIP, such as the Beale-Orchard-Hays Prize. From 2006, Dr. Achterberg worked for ILOG / IBM as developer of CPLEX in versions 11 to 12.6. Since 2014 he participates in the development of the Gurobi Optimizer, currently being the Vice President of R&D at Gurobi. In his spare time, Dr. Achterberg continues to work on the Gurobi MIP solver, because he is addicted to MIP. He does not have any interest in other things; his spouse, his three children, his two drum kits, and his attendances in the moshpits of punk rock concerts are just mock-ups to pretend having a normal life. Contents of the video: 0:00 - Intro 1:46 - Family background 4:59 - Cold war 7:50 - History classes 9:44 - Brain surgery that affected his vision 13:14 - Drumming and Punk Rock 15:46 - Learning to code in Basic at a very young age 18:03 - Masters degree in Mathematics and in Computer Science 19:05 - Learning about Optimization and implementing the Simplex algorithm 21:47 - Masters dissertation in Mathematics 23:13 - Masters dissertation in Computer Science (reliability branching) 26:10 - Creating the SCIP framework during the PhD 28:43 - Combining Constraint Programming with MIP 30:32 - Conflict analysis 33:04 - Is SCIP fully open-source? 35:26 - Joining ILOG 40:00 - Main contributions to the CPLEX solver 42:36 - Joining Gurobi 44:30 - Working with the Gurobi co-founders 46:30 - Main contributions to the Gurobi solver 48:56 - "Computational work is not appreciated as much at it should be" in Academia 53:37 - Branching tips 57:03 - Primal heuristics 59:51 - Automatic decomposition 1:02:34 - Is it worth developing a commercial branch-cut-and-price solver? 1:05:05 - Defining default parameter values for a solver 1:06:27 - How to extract the best from the solver? 1:08:53 - Understanding the meaning of feasibility tolerance in MIP solvers 1:10:58 - Why commercial MIP solvers do not support GPU? 1:13:03 - How can Constraint Programming enhance the performance of MIP solvers? 1:16:13 - Using Machine Learning to improve the performance of MIP solvers 1:17:42 - Employing generative AI to make the use of optimization tools more accessible 1:20:14 - The advantage of not to be so aware of what is out there 1:22:30 - Life without MIP? 1:23:30 - Concluding remarks
Dr. Achterberg studied mathematics and computer science at the Technical University of Berlin and the Zuse Institute Berlin. He finished his PhD in mathematics under supervision of Prof. Martin Grötschel in 2007. Dr. Achterberg is the author of SCIP, currently the best academic MIP solver. In addition to numerous publications in scientific journals he has also received several awards for his dissertation and for SCIP, such as the Beale-Orchard-Hays Prize. From 2006, Dr. Achterberg worked for ILOG / IBM as developer of CPLEX in versions 11 to 12.6. Since 2014 he participates in the development of the Gurobi Optimizer, currently being the Vice President of R&D at Gurobi. In his spare time, Dr. Achterberg continues to work on the Gurobi MIP solver, because he is addicted to MIP. He does not have any interest in other things; his spouse, his three children, his two drum kits, and his attendances in the moshpits of punk rock concerts are just mock-ups to pretend having a normal life. Contents of the video: 0:00 - Intro 1:46 - Family background 4:59 - Cold war 7:50 - History classes 9:44 - Brain surgery that affected his vision 13:14 - Drumming and Punk Rock 15:46 - Learning to code in Basic at a very young age 18:03 - Masters degree in Mathematics and in Computer Science 19:05 - Learning about Optimization and implementing the Simplex algorithm 21:47 - Masters dissertation in Mathematics 23:13 - Masters dissertation in Computer Science (reliability branching) 26:10 - Creating the SCIP framework during the PhD 28:43 - Combining Constraint Programming with MIP 30:32 - Conflict analysis 33:04 - Is SCIP fully open-source? 35:26 - Joining ILOG 40:00 - Main contributions to the CPLEX solver 42:36 - Joining Gurobi 44:30 - Working with the Gurobi co-founders 46:30 - Main contributions to the Gurobi solver 48:56 - "Computational work is not appreciated as much at it should be" in Academia 53:37 - Branching tips 57:03 - Primal heuristics 59:51 - Automatic decomposition 1:02:34 - Is it worth developing a commercial branch-cut-and-price solver? 1:05:05 - Defining default parameter values for a solver 1:06:27 - How to extract the best from the solver? 1:08:53 - Understanding the meaning of feasibility tolerance in MIP solvers 1:10:58 - Why commercial MIP solvers do not support GPU? 1:13:03 - How can Constraint Programming enhance the performance of MIP solvers? 1:16:13 - Using Machine Learning to improve the performance of MIP solvers 1:17:42 - Employing generative AI to make the use of optimization tools more accessible 1:20:14 - The advantage of not to be so aware of what is out there 1:22:30 - Life without MIP? 1:23:30 - Concluding remarks