Univerza na Primorskem Fakulteta za matematiko, naravoslovje in informacijske tehnologije
SI | EN

Ponedeljkov seminar računalništva in informatike - Arhiv

2024 2023 2022 2021 2020 2019 2018 2017 2016 2015 2014 2013 2012
1 2 3 4 5 6 7 8 9 10 11 12

ponedeljek, 20. junij 2022 Nedim ŠIŠIĆ: Neural Network Based Classification of Brain Lesions

V ponedeljek, 20. junija 2022, bo ob 16.00 uri prek spletnih orodij na daljavo izvedeno
predavanje v okviru PONEDELJKOVEGA SEMINARJA RAČUNALNIŠTVA IN INFORMATIKE
Oddelkov za Informacijske znanosti in tehnologije UP FAMNIT in UP IAM.

ČAS/PROSTOR: 20. junij 2022 ob 16.00 na daljavo

-------------------------------------
PREDAVATELJ: Nedim ŠIŠIĆ
-------------------------------------

Nedim Šišić is a FAMNIT student from Bosnia and Herzegovina. He is enrolled in the 2nd year of the Computer Science Master's study program. His Research Seminar mentor is Professor Peter Rogelj.

-------------------------------------------------------------------------------------
NASLOV: Neural Network Based Classification of Brain Lesions
-------------------------------------------------------------------------------------

POVZETEK:

A brain lesion is an area of tissue injury or disease within the brain. Brain-imaging techniques, such as MRI, produce images of the brain used for diagnosis or prognosis of a wide variety of injuries or conditions. Such images can be used to classify brain lesions. Recently, neural networks have been used to great success in different computer vision domains, including brain image processing. In this work, we develop a convolutional neural network that automatically classifies MRI images in multiple sclerosis patients. The image dataset consists of three-dimensional images with three channels: FLAIR MRI, QSM MRI, and a mask channel indicating lesion location. We describe data augmentation techniques we use for increasing the dataset and present the network architecture and the results of the model. Specifically, we show the results of training the network using different combinations of the three image channels.

---------------------------------------------------------------------------------------------------

Seminar bo potekal v angleškem jeziku v Zoom "klepetalnici" na naslovu:

https://upr-si.zoom.us/j/297328207?pwd=S3Zpdk1VR3pjckNtWkQwKzlvcDR5UT09

Vabljeni! 


ponedeljek, 13. junij 2022 Martin JAKOMIN, Blaž MRAMOR: Data science at Zemanta

V ponedeljek, 13. junija 2022, bo ob 16.00 uri prek spletnih orodij na daljavo izvedeno
predavanje v okviru PONEDELJKOVEGA SEMINARJA RAČUNALNIŠTVA IN INFORMATIKE
Oddelkov za Informacijske znanosti in tehnologije UP FAMNIT in UP IAM.

ČAS/PROSTOR: 13. junij 2022 ob 16.00 na daljavo

--------------------------------------------------------------------------------
PREDAVATELJA: Martin JAKOMIN, Blaž MRAMOR (Zemanta)
--------------------------------------------------------------------------------

Martin Jakomin received his PhD in Computer science from University of Ljubljana in 2019. His research interests include fields of data science and machine learning. More specifically, he is interested in the areas of recommender systems, incremental learning, data fusion and data generation.
Blaž Mramor obtained his PhD in Mathematics from VU University Amsterdam in 2012 and worked afterwards as a researcher in the fields of Geometry and Analysis at University of Freiburg and at UP FAMNIT. Since 2017 he has been working as a data scientist and ML engineer, first in Allianz SE in Munich and since 2021 in Zemanta.

-----------------------------------------------
NASLOV: Data science at Zemanta
-----------------------------------------------

POVZETEK:

We will first introduce Zemanta's line of business, real-time bidding, which is an automated auction process for the purchase of on-line ad spaces. Then we will briefly present the different data science areas and the challenges that we are facing. Finally, we will focus our attention on one of our major components, the click-through-rate predictor, and explain some details about the data, the training pipelines and the choice of our ML algorithms for it.

---------------------------------------------------------------------------------------------------

Seminar bo potekal v angleškem jeziku v Zoom "klepetalnici" na naslovu:

https://upr-si.zoom.us/j/297328207?pwd=S3Zpdk1VR3pjckNtWkQwKzlvcDR5UT09

Vabljeni!