5 PyData Piraeus meetup: SOTA DNNs& Graph-based Recommender Systems

Date(s) - 28/02/2020
19:00 - 21:00

New York College

Κατηγορία(ες) Δεν υπάρχουν κατηγορίες

Hello all,

The next PyData meetup will take place at NYC on Friday, 28th February at 7PM. We are privileged to welcome Dr. Dimitrios Oikonomou, Co-Founder and CTO at Earth Science Analytics AS, Norway, who will talk about SOTA Deep Neural Networks on limited training datasets and Ilias Tsoumas, Computer Scientist and Affiliated researcher at University of Piraeus, who will introduce us to graph-based Recommender Systems and Latent Factor Models.

We would like to thank our Sponsors, wappier, New York College and NumFocus.

At the end of the meetup, refreshments will be served.

This event is dedicated to the memory of my father, Spiridon, who supported me every step of the way. We will always keep you in our hearts!

[masked] SOTA DNNs: How good are they, and how do we rank their performance when a limited training set is available?, Dr. Dimitrios Oikonomou is cofounder and CTO at Earth Science Analytics, Norway.

Deep learning algorithms have been used to solve several imaging tasks such as classification and segmentation. Such problems target to classify each pixel of an image into a class. Each class, corresponds to some identifiable unit in the image belonging to different categories.

In this talk, we’ll review the architecture of a selection of well known, state-of-the-art (SOTA) Deep Neural Networks (DNN) models ; Segnet, Unet, PSPnet, DeepLab as well as some new architectures that have been adapted from the state of the art ones by reducing the number of layers and filters in them targeting reduced overfitting and regularisation.

We target to understand how the SOTA architectures performs when limited training sets are available and elaborate on techniques that could improve their performance.

Examples will be given using paradigms from seismic interpretation.

Dr. Dimitrios Oikonomou is cofounder and CTO of Earth Science Analytics AS, Norway. Has served as CTO and various management board positions in defense industry for more than 10 years, was R&D team leader and member of strategy committee in Emerson Roxar. He has extensive experience from software development, programming, and machine learning for more than 20 years. Inventor of two patents, one of them internationally registered by Emerson Roxar AS as a sole inventor. He holds a PhD in electromagnetics and has more than 25 publications in peer reviewed international journals.

LinkedIn: https://www.linkedin.com/in/dr-dimitrios-oikonomou-297121/
Twitter: @OikonomouDimi
Earth Science Analytics Website: https://earthanalytics.ai/
[masked] Intro to Recommender Systems & A graph-based approach, Ilias Tsoumas, Computer Scientist, Affiliated researcher at University of Piraeus

This presentation consists of two parts. We begin by introducing the logic behind recommender systems. Specifically, the aspects covered regard the traditional recommender systems, intersection of them with link prediction and beyond accuracy metrics. In the second part we present an approach which combines latent factor models with graph-based models towards microservices recommendations, constrained by the small(ish) data of a real case.

Ilias Tsoumas is a curious person for whom learning and innovation have always played an important role. He has been engaged in computer science and mostly in field of machine learning, in particular he has dealt more with recommender systems, clustering and nlp. He has worked for three years on research and innovation projects as researcher under the Daclab at University of Piraeus with focus on 5G and its intersection with machine learning. Currently, he is completing his master degree in Data Science at Athens University of Economics and Business. He enjoys to involve in problem solving cases, especially, related to his field.

LinkedIn: https://www.linkedin.com/in/itsoum/
University of Piraeus Data& Cloud Laboratory Website: https://daclab.ds.unipi.gr/

See you all at NYC!

Όλες οι εκδηλώσεις