Talks Old

Albert S. Berahas/Talks Old

To create today is to create dangerously. Any publication is an act, and that act exposes one to the passions of an age that forgives nothing.

Albert Camus

2021
  • (upcoming) EURO 2021, 31st European Conference on Operations Research, Athens, Greece.
    • Analysis of Line Search and Trust Region Methods with Noise
  • (upcoming, virtual) 3rd IMA and OR Society Conference on Mathematics of Operational Research.
    • Sequential Quadratic Optimization for Nonlinear Equality Constrained Stochastic Optimization
  • (upcoming, virtual) CSE 2021, SIAM Conference on Computational Science and Engineering.
    • A Symmetric Blockwise Truncated Optimization Algorithm for Machine Learning
2020
  • (virtual) INFORMS Annual Meeting 2020, Institute for Operations Research and the Management Sciences Conference, National Harbor, MD.
    • Analysis Of A Generic Line Search Method With Noise: Algorithms, Convergence Rates And Gradient Approximation. (slides)
  • (postponed due to COVID-19) DataX Workshop (September 2020): Old and New Open Questions in Optimization, Princeton University, Princeton, NJ.
    • Limited-Memory BFGS with Displacement Aggregation.
  • (postponed due to COVID-19) SIOPT 2020, SIAM Conference on Optimization, Hong Kong.
    • Global Convergence Rate Analysis of a Line Search Algorithm with Noise.
  • (postponed due to COVID-19) MDS 2020: SIAM Conference on Mathematics of Data Science, Cincinnati, OH.
    • Sampled Quasi-Newton Methods For Deep Learning.
  • JMM 2020, Joint Mathematics Meetings, Denver, CO.
    • A Theoretical and Empirical Comparison of Gradient Approximations in Derivative-Free Optimization: Algorithms, Convergence Analysis and Noise. (slides)
2019
  • Numerical Analysis and Scientific Computing Seminar, November 2019, Courant Institute of Mathematical Sciences, New York University, New York, NY. 
    • Limited-Memory BFGS with Displacement Aggregation. (slides)
  • INFORMS Annual Meeting 2019, Institute for Operations Research and the Management Sciences Conference, Seattle, WA.
    • Quasi-Newton Methods For Deep Learning: Forget The Past, Just Sample. (slides) 
  • Young Researchers Workshop 2019, ORIE, Cornell University, Ithaca, NY. 
    • Global Convergence Rate Analysis of a Generic Line Search Algorithm with Noise. (slides)
  • ICCOPT 2019, International Conference on Continuous Optimization, Berlin, Germany.
    • Aggregated Quasi-Newton Methods. (slides)
  • INFORMS PhD Seminar 2019, Lehigh University, Bethlehem, PA.
    • A Theoretical and Empirical Comparison of Gradient Approximations in Derivative-Free Optimization. (slides)
  • CSE 2019, SIAM Conference on Computational Science and Engineering, Spokane, WA.
    • Scaling up and Randomizing Derivative-Free Optimization for Machine Learning. (slides)
2018
  • INFORMS Annual Meeting 2018, Institute for Operations Research and the Management Sciences Conference, Phoenix, AZ.
    • Stochastic Quasi-Newton Methods – Past, Present and Future. (slides)
  • 2nd NSF TRIPODS PI Workshop 2018,  University of California, Santa Cruz, CA
    • Do we Need 2nd Order Methods in Machine Learning?  (poster)
  • MOPTA 2018, Modeling and Optimization: Theory and Applications, Bethlehem, PA.
    • Derivative-Free Optimization of Noisy Functions via Quasi-Newton Methods. (slides)
  • ISMP 2018, International Symposium on Mathematical Programming, Bordeaux, France.
    • Derivative-Free Optimization of Noisy Functions via Quasi-Newton Methods. (slides)
  • SPIE-MRSEC Student Seminar Series 2018, Northwestern University, Evanston, IL.
    • Methods for Large Scale Nonlinear and Stochastic Optimization. (slides)
2017
  • OptML Seminar 2017, Lehigh University, Bethlehem, PA.
    • Stochastic Newton and Stochastic Quasi-Newton Methods for Nonlinear Optimization. (slides)
  • INFORMS Annual Meeting 2017, Institute for Operations Research and the Management Sciences Conference, Houston, TX.
    • Derivative-Free Optimization via Quasi-Newton Methods. (slides) 
  • IFORS 2017, International Federations on Operations Research Societies Conference, Quebec city, Canada.
    • Balancing Computation and Communication in Distributed Optimization. (slides)
  • MMLS 2017, Midwest Machine Learning Symposium, Chicago, IL.
    • Are Newton-Sketch and Subsampled Newton Methods Effective in Practice? (poster)
  • SIOPT 2017, SIAM Conference on Optimization, Vancouver, Canada.
    • A Numerical Investigation of Sketching and Subsampling. (slides)
  • ACNTW 2017, Northwestern University, Chicago, IL.
    • Are Newton-Sketch and Subsampled Newton Methods Effective in Practice? (poster)
  • CASSC 2017, SIAM Student Conference (Chicago area), Evanston, IL.
    • A Multi-Batch L-BFGS Method for Machine Learning. (slides)
2016
  • NIPS 2016, Neural Information Processing Systems Conference, Barcelona, Spain.
  • ECML 2016, European Conference on Machine Learning, Riva del Garda, Italy.
    • adaQN: An Adaptive Quasi-Newton Algorithm for Training RNNs. (slides, poster)
  • Fields Institute Workshop on Nonlinear Optimization Algorithms and Industrial Applications 2016, Toronto, Canada.
    • A Robust Multi-Batch L-BFGS Method for Machine Learning. (poster)
  • CASSC 2016, SIAM Student Conference (Chicago area), Chicago, IL. 
    • An Adaptive Quasi-Newton Algorithm for Training RNNs. (slides)