BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:America/New_York BEGIN:DAYLIGHT DTSTART:20250309T030000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3 TZNAME:EDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20241103T010000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11 TZNAME:EST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20250306T000400Z UID:372B70C2-0160-49A0-B81C-0AEEB6678F90 DTSTART;TZID=America/New_York:20250305T160000 DTEND;TZID=America/New_York:20250305T170000 DESCRIPTION:Abstract: First-order methods (FOMs) are optimization algorithm s that can be described and analyzed using the values and (sub)gradients o f the functions to be minimized. FOMs are the main workhorses for modern l arge-scale optimization and machine learning due to their low iteration co sts\, minimal memory requirements\, and dimension-independent convergence guarantees. As the data revolution continues to unfold\, the pressing dema nd for faster FOMs has become the key problem in today’s big data era. T o that goal\, we present Branch-and-Bound Performance Estimation Programmi ng (BnB-PEP)\, the first unified methodology for constructing optimal (pro vably fastest) FOMs for convex and nonconvex optimization. BnB-PEP poses t he problem of finding the optimal FOM as a nonconvex but practically tract able quadratically constrained quadratic optimization problem and solves i t to certifiable global optimality using a custom branch-and-bound algorit hm. Our open-source custom branch-and-bound algorithm\, through exploiting specific problem structures\, outperforms the latest off-the-shelf softwa re by orders of magnitude\, accelerating the solution time to discover the optimal FOMs from hours to seconds and weeks to minutes. We apply BnB-PEP to several practically relevant convex and nonconvex setups and obtain FO Ms with bounds that improve upon prior state-of-the-art results. Furthermo re\, we use the BnB-PEP methodology to find proofs with potential function structures\, thereby systematically generating analytical convergence pro ofs. Recently\, BnB-PEP has helped pave the way for some fundamental resul ts in optimization: (i) novel momentum-free accelerated gradient descent m ethods that broke with decades of conventional wisdom (ii) the optimal FOM for constrained convex optimization that outperforms the celebrated FISTA algorithm by Beck and Teboulle\, and (iii) the first non-asymptotic conve rgence theory for the widely used nonlinear conjugate gradient methods. (j oint work with Bart Van Parys and Ernest K. Ryu)\n\nPaper link: https://ar xiv.org/pdf/2203.07305\n\nSpeaker(s): Shuvomoy Das Gupta\n\nAgenda: \n- Ta lk by Shuvomoy Das Gupta at 4:00 pm\n- Dinner box after the talk at 5:00 p m\n- You don't have to be an IEEE member to attend this meeting.\n\nRoom: 202\, Bldg: ECE\, 141 Warren St\, New Jersey Institute of Technology\, New ark\, New Jersey\, United States\, 07103 LOCATION:Room: 202\, Bldg: ECE\, 141 Warren St\, New Jersey Institute of Te chnology\, Newark\, New Jersey\, United States\, 07103 ORGANIZER:marcos.netto@njit.edu SEQUENCE:22 SUMMARY:Branch-and-Bound Performance Estimation Programming: A Unified Meth odology for Constructing Optimal Optimization Methods URL;VALUE=URI:https://events.vtools.ieee.org/m/471375 X-ALT-DESC:Description: <br /><div><strong>Abstract:</strong> First-order m ethods (FOMs) are optimization algorithms that can be described and analyz ed using the values and (sub)gradients of the functions to be minimized. F OMs are the main workhorses for modern large-scale optimization and machin e learning due to their low iteration costs\, minimal memory requirements\ , and dimension-independent convergence guarantees. As the data revolution continues to unfold\, the pressing demand for faster FOMs has become the key problem in today&rsquo\;s big data era. To that goal\, we present Bran ch-and-Bound Performance Estimation Programming (BnB-PEP)\, the first unif ied methodology for constructing optimal (provably fastest) FOMs for conve x and nonconvex optimization. BnB-PEP poses the problem of finding the opt imal FOM as a nonconvex but practically tractable quadratically constraine d quadratic optimization problem and solves it to certifiable global optim ality using a custom branch-and-bound algorithm. Our open-source custom br anch-and-bound algorithm\, through exploiting specific problem structures\ , outperforms the latest off-the-shelf software by orders of magnitude\, a ccelerating the solution time to discover the optimal FOMs from hours to s econds and weeks to minutes. We apply BnB-PEP to several practically relev ant convex and nonconvex setups and obtain FOMs with bounds that improve u pon prior state-of-the-art results. Furthermore\, we use the BnB-PEP metho dology to find proofs with potential function structures\, thereby systema tically generating analytical convergence proofs. Recently\, BnB-PEP has h elped pave the way for some fundamental results in optimization: (i) novel momentum-free accelerated gradient descent methods that broke with decade s of conventional wisdom (ii) the optimal FOM for constrained convex optim ization that outperforms the celebrated FISTA algorithm by Beck and Teboul le\, and (iii) the first non-asymptotic convergence theory for the widely used nonlinear conjugate gradient methods. (joint work with Bart Van Parys and Ernest K. Ryu)</div>\n<div>&nbsp\;</div>\n<div><strong>Paper link:</s trong>&nbsp\;<a href="https://arxiv.org/pdf/2203.07305" target="_blank" re l="noopener" data-saferedirecturl="https://www.google.com/url?q=https://ar xiv.org/pdf/2203.07305&amp\;source=gmail&amp\;ust=1740663951247000&amp\;us g=AOvVaw0SslotL03WSPKs9PELGnAZ">https://arxiv.org/pdf/<wbr>2203.07305</a>< /div><br /><br />Agenda: <br /><p>- Talk by Shuvomoy Das Gupta at 4:00 pm< br>- Dinner box after the talk at 5:00 pm<br>- You don't have to be an IEE E member to attend this meeting.</p> END:VEVENT END:VCALENDAR