Noticimat 15

Seminarios

Seminario de Computación 

Hora: 12:30 (GMT-6)

Enlace de transmisión: https://bluejeans.com/789853847

Ponente: Dr. Alejandro Rosales Pérez (CIMAT, Monterrey)

Título: Optimización bayesiana evolutiva: ¿qué es y para qué?

Resumen: Los algoritmos evolutivos son técnicas poderosas para optimización de problemas de caja negra; sin embargo, pueden llegar a requerir de muchas evaluaciones para aproximarse a una solución buena. La optimización Bayesiana es una alternativa para tratar con problemas cuyas evaluaciones son computacionalmente costosas. En esta charla, revisaremos los principios básicos de estas dos estrategias y cómo se combinan. También, se presentarán algunas aplicaciones dentro del área de aprendizaje máquina y cerraremos con algunas líneas de investigación potenciales.

 

Hora:  12:30 (GMT-6)

Enlace de transmisión: https://bluejeans.com/508629789 

Ponente: Julio Enrique Castrillon Candas (Boston University)

Título: A stochastic functional (data) analysis perspective to anomaly detection and robust machine learning 

Resumen: In this talk we discuss the connections between stochastic functional (data) analysis, anomaly detection and machine learning. Based on tensor product representations of random fields and stochastic processes, a series of orthogonal nested subspaces is constructed for detecting anomalous signal components.  In particular this can succeed even if the stochastic behavior of the signal changes either in a global or local sense. This approach is applicable to general topologies, including geo-spatial, spatio-temporal, manifold, network, etc., or any combination of these. The construction of nested subspaces involves efficient algorithms from computational applied mathematics and high performance computing. A mathematical approach is developed to locate and measure sizes of extraneous components based on construction of multilevel nested subspaces. We construct an effective hypothesis test depending on the degree of the multilevel spaces and decay of the eigenvalues of the covariance function.  No assumptions on independence nor on the distribution of the data are needed. We illustrate applications to deforestation in remote sensing, and machine learning cancer diagnosis. This method leads to a significant increase in precision and accuracy over the current top benchmarks for the Global Cancer Map gene expression network dataset. Joint work with Mark Kon.

 

Hora: 09:00 (GMT-6)

Enlace de transmisión: https://bluejeans.com/387784372

Ponente: Ilia Itenberg (IMJ-PRG Paris)

Título: Planes in cubic fourfolds

Resumen: We discuss possible numbers of 2-planes in a smooth cubic hypersurface in the 5-dimensional projective space. We show that, in the complex case, the maximal number of planes is 405, the maximum being realized by the Fermat cubic. In the real case, the maximal number of planes is 357. The proofs deal with the period spaces of cubic hypersurfaces in the 5-dimensional complex projective space and are based on the global Torelli theorem and the surjectivity of the period map for these hypersurfaces, as well as on Nikulin's theory of discriminant forms. Joint work with Alex Degtyarev and John Christian Ottem.

Web del seminario: https://researchseminars.org/seminar/LAGARTOS 

 

Hora: 10:00 (GMT-6)

https://us06web.zoom.us/j/81057858436?pwd=cDNBQlBFbGgyWHMxT05vWkxJZ2ttUT

09

Ponente: Alexander Grosberg (New York University)

Título: Is Trivial Knot Really So Trivial?

Resumen: While topological ideas are widely popular in physics, topology of classical linear threads of polymers presents steep mathematical and conceptual challenges, with applications in both biopolymers and materials. I will concentrate on the simplest case of polymer unknots and review what is known about fluctuations and statistical mechanics of such objects based mostly on simulations, experiments, and hand-waving theoretical arguments. Continuing with increasingly sophisticated models and phenomena, I will review several more recent theoretical and experimental achievements, and conclude with the discussion of a controversial concept of ''topological glass''.

Web del seminario: https://seminargeotop-a.com/ 

 

Divulgación

Se llevará a cabo el segundo Divertimate virtual. Divertimate es una feria virtual organizada por los grupos de divulgación de las diferentes unidades del CIMAT (Centro de Investigación en Matemáticas) en la que podrás interactuar con comunidad matemática, convivir con otros curiosos de las matemáticas, conocer cómo se ven las unidades del CIMAT y explorar distintas salas en donde habrá juegos, talleres, acertijos, exposiciones, charlas y más. 

Horario: 10:00 a 13:00 horas (horario de la Ciudad de México) 

Plataforma: Gather Town

 

Valentina Muñoz Porras, Carmen Mares Orozco, Dennis Olvera Torres y Marco Figueroa Ibarra ofrecerán clases de matemáticas a las niñas y niños de la Escuela Primaria "Amado Nervo", ubicada en la colonia Valenciana, Guanajuato, 

Gto. Horario: 8:30 a 10:30 am.