[Todos] CHARLA IFLYSIB EXTRAORDINARIA - Miércoles 21/08/2024, 10:30hs - "Detección no supervisada de correlaciones semánticas en redes neuronales artificiales"

[Charlas IFLYSIB]

*¡¡DÍA EXTRAORDINARIO!!*


*Charla  IFLYSIB (en castellano)*
*MIÉRCOLES 21/08/2024, 10:30 hs.*
*Lugar: IFLySiB (59 #789 e/10 y 11, La Plata)*

*La charla será presencial pero **también** será transmitida vía Zoom en:*
https://utn.zoom.us/j/89038134365


*Título:*
*Detección no supervisada de correlaciones semánticas en redes neuronales
artificiales*


*Expositor: *

*Santiago Acevedo - International School for Advanced Studies (SISSA),
Trieste, Italia*

*Resumen: *
*In real-world data, information is stored in extremely large feature
vectors. These features are often correlated due to complex interactions
involving multiple features simultaneously. These correlations
qualitatively correspond to semantic roles and are naturally recognized by
both the human brain and artificial neural networks. This recognition
enables, for instance, the prediction of missing parts of an image or text
based on their context. We present a method to detect such correlations in
high-dimensional data, represented as binary numbers. We estimate the
binary intrinsic dimension of a dataset, which quantifies the minimum
number of independent coordinates needed to describe the data, and is
therefore a proxy of semantic complexity. The proposed algorithm is largely
insensitive to the so-called curse of dimensionality, and can therefore be
used in big data analysis. We test this approach identifying phase
transitions in model magnetic systems and we then apply it to the detection
of semantic correlations of images and text inside deep neural networks.*




-- 
*Comisión ChIFLy*
______________________________________________

Si no tenés interés en recibir los avisos de las chiflys, escribinos.
------------ próxima parte ------------
Se ha borrado un adjunto en formato HTML...
URL: <http://mail.fisica.unlp.edu.ar/pipermail/todos/attachments/20240819/8a9f55ec/attachment.html>