Deep Learning In Vision / NLP / Time Series Prediction

Kindly remind you that the day of event has been changed to 20th due to some speakers availability. Again please make sure you can attend before registering.

We are going to host our 4th meetup this year in Travelport, and this is our first time to have the event in their venue.

I am so excited to bring 3 speakers from universities to show their projects on the day. Please keep checking this page we will have more details updated soon.

*** AGENDA ***

[18:00] Registration & Social Networking, enjoy the food and drinks

[18:30]
Welcome and opening by Stephen Oman, Director of Data Analytics at Travelport, he will also give us a speech to introduce the company and the current projects his team are working on.

[18:45 - 19:15]
By Phd Mr. Linyi Yang from UCD INSIGHT

Title: BERT: The Milestone Technique for NLP

Abstract:
BERT was developed by Google AI Language and came out in Oct. 2018. It has achieved the best performance in many NLP tasks. Today, if you say you don’t hear about BERT, you can not say you are the domain expert in NLP. In the past, the best NLP model would be the GRU+Attention architecture, and the best pre-training technique for NLP include word2vec (2013), glove (2014), Elmo (2017). However, at the end of 2018, Google open-sourced a new technique for NLP pre-training called Bidirectional Encoder Representations from Transformers, or BERT, it sets new standards in 11 Language Tasks.

[19:15 - 20:00]
By Professor Aljosa Smolic and his team,
Phd Sebastian Lutz
Phd Koustav Ghosal
Phd Tejo Chalasani

Title: Deep Learning for Visual Computing in V-SENSE

Abstract:
Artificial Intelligence (AI) has made it from science fiction into everyday life. Machine Learning (ML) enabled breakthroughs due to availability of massive data and computational resources. Deep Learning (DL) in particular disrupted all areas of visual computing (VC), including computer vision/graphics and image/video processing. The V-SENSE team of Trinity College Dublin adopted this challenge and opportunity, and made a number of significant contributions to the field of DL for VC over the last 2 years, which were published at different venues. Prof Smolic will give an introduction and highlight some of those, including deep normal estimation, DeepTMO for HDR tone mapping, SalNet360 for 360 image saliency estimation, and others.

[20:00 - 20:30]
By Phd Mr. Fouad Bahrpeyma from DCU INSIGHT

Title: Multi-Step Ahead Time Series Prediction

Abstract:
Where the goal is to provide accurate and reliable forecasts over long horizons.
Among our contributions to the field I can highlight the following two:
1- We've introduced a new multi step ahead prediction strategy known as Multi-Resolution Forecast Aggregation (MRFA).
2- We've proposed a new meta-analysis method for enabling non-experts to pick the right time series prediction method.
We've also provided a benchmarking method to assist the community with advanced and featured testing data.

A wide variety of phenomena are characterized by time dependent dynamics that can be analyzed using time series methods. Various time series analysis techniques have been presented, each addressing certain aspects of the data. In time series analysis, forecasting is a challenging problem when attempting to estimate extended time horizons which effectively encapsulate multi-step-ahead (MSA) predictions. Two original solutions to MSA are the direct and the recursive approaches. Recent studies have mainly focused on combining previous methods as an attempt to overcome the problem of discarding sequential correlation in the direct strategy or accumulation of error in the recursive strategy. We introduce introduce Multi-Resolution Forecast Aggregation (MRFA) which incorporates Resolutions of Impact to improve MSAP performance.

[20:30 - 21:00] Networking till the event close