Project Demos & Panel Discussion In IR / NLP / ANN

Location image of event venue


We only accept 65 people for this event.
Please do not register the event if you are not coming.

The event has also been published on

The first time, Deep Learning Dublin will hold our meetup @ DELOITTE DIGITAL. And also the first time, we involve a panel discussion stage.

Panel Discussion Topic: Machine vs. Human in Information Cognition

*** AGENDA ***

[18:15] Registration & Social Networking, enjoy the Pizza and drinks

[19:00] Welcome and opening by Deloitte Digital Engineering Head Alan Liu, and he will also give us a speech to introduce the company and the current projects his team are working on.

[19:15 - 19:40]
By Phd Eoin Kenny from UCD INSIGHT

Twin-Systems to Explain Artificial Neural Networks Using Case-Based Reasoning

In this presentation, our recent publication is summarized which addresses the eXplainable AI (XAI) problem. The talk details twin-systems, where a black box model is mapped to a white box “twin” that is more interpretable, with both systems using the same dataset. The framework is instantiated by twinning an artificial neural network (ANN; black box) with a case-based reasoning system (CBR; white box), and mapping the feature weights from the former to the latter to find cases that explain the ANN’s outputs. Using a novel evaluation method, the effectiveness of this twin-system approach is demonstrated by showing that nearest neighbor cases can be found to match the ANN predictions for benchmark datasets. Several feature-weighting methods are competitively tested in two experiments, including a novel, contributions-based method (called COLE) that is found to perform best. The tests consider the "twinning" of traditional multilayer perceptron (MLP) networks and convolutional neural networks (CNN) with CBR systems. For the CNNs trained on image data, qualitative evidence shows that cases provide plausible explanations for the CNN's classifications.

[19:40 - 20:05]
By Dr. Wei Li, from DCU ADAPT

IR Methodologies

The presentation will introduce the general idea of Information Retrial(IR) system. Including the basic workflow of IR system, the challenge we are facing and few real word projects will be discussed.
The example case show
1) how does NLP method employed in search.
2) The different indexing methods
3) The search and evaluation strategy performed.

[20:05 - 20:30]
By Dr. Michael Scriney, from DCU INSIGHT

Ghost Buses: Anomaly detection of transport data.

Transport providers supply real time information streams to consumers concerning the status of their network.

These streams generally manifest themselves as a measure of time indicating when a bus/train/luas etc.. arrives at a specific transportation point.

However, there can be a disconnect between what the real-time system displays and the status of the network in the real-world.

The purpose of this work is to analyse these streams using deep learning in order to build a classifier which can mark anomalies coming from the real-time systems in order to provide consumers with more information.

[20:30 - 21:00] Panel Discussion
Title: Machine vs. Human in Information Cognition

Speech given By: Dr. Caitríona Osborne, UCD
Title: Human Translation Methodologies
1. Concept of language and culture inherently linked
a. Sapir-Whorf hypothesis:
i. cultural differences are articulated in language;
ii. how we view the world is largely determined by our thought processes;
iii. language shapes how we think and see the world.
b. Can a machine pick up on cultural inferences?
2. What does ‘translation’ mean?
3. Domestication and foreignization
4. Some tactics used by translators include
5. Activities/ examples

Discussion Involves all speakers and audience.

[21:00 - 21:30] Social Networking Until Event Close
[21:30] Event Close