MSc Project Showcase - North East Data Meetup, September 2022
The Aberdeen Data Meet-up is back!
The season begins with one of our most popular event formats - The MSc Student Showcase where we will have six 10-minute presentations from students in the 2021-2022 MSc Data Science Cohort at local universities speaking to us about their summer project.
Expect a mix of industry and academia, technical and business people coming together to network, listen to informative talks, share news, and discuss collaborations.
This is a hybrid event, and we welcome in-person attendees to join us from 6pm for pizza, drinks, and networking. Online attendees can join from 6.15pm, with presentations starting at 6.30pm.
Numbers are limited, so register today!
If you wish to attend online, please book a place AND email email@example.com to confirm and a link will be sent to you.
Talks (in no particular order):
1. Sree Raghav Duvvuri - Facial Emotion Recognition With Augmented Dataset and Evaluation
Precis: This project deals aims to build a neural network for facial emotion recognition to classify an image into one of the seven emotions (angry, fear, happy, neutral, sad, disgust and surprise) and evaluate its performance with respect to a fully trained ResNet50 model and another model, which is a RessNet50 model, in which, only the first 101 layers are trained and the rest use weights from imagenet.
Bio: Sree is a student pursuing MSc in Artificial Intelligence at the University of Aberdeen, eager to break into the industry.
2. Kavyasree Ande - Anomaly Detection with GAN stabilisation for Smoke Dataset
Anomaly detection has major application is industry, to avoid major industrial accidents and mishaps. In recent times much research has been done on deep generative models for anomaly detection, as generative models can learn only from standard data, and there is still ongoing research on improving the methods to maximize the model performance.
In this project, experiments were designed to investigate DCGAN and AnoGAN architectures in depth and a pipeline for applying AnoGAN to smoke data set with stable GAN training has been achieved.
Kavyasree is an MSc Artificial Intelligence student at University of Aberdeen.
3. Ali Hasan - Assessment of Scottish Open Data Policies & Implementation - A Data User Perspective
Ali's talk will provide an overview of the open data movement in Scotland and highlight improvement areas in policy and implementation. The underlining argument is that data users must be the focus when policies are devised and implemented since it is they who derive socio-economic value from open data.
Ali is pursuing an MSc Business Analytics at RGU, and is interested in socio-economic value of open data and all things open in general.
4. Prashasthi Narendra Gowdar - Semantic Alignment In Dialogue
The project aims to understand the patterns in human conversation as the dialogue develops, speakers have been observed to adapt to their dialogue partner at several levels of communication, including the phrases they use, how quickly they talk, and their style as well as manner of speech. The primary goal of the research is to observe convergence or divergence patterns on a semantic scale using various neural language models in order to determine whether a linguistic theory known as the Interactive Alignment Model (IAM) is true in all cases.
Prashasthi is an MSc Artificial Intelligence student at the University of Aberdeen.
5. Rahul Baburajan: A Creative Writing Collaborative Tool using Large Language Models.
The project attempts to model a conditional language generation model using GPT-Neo that can generate text in the writing style of Robert Burns, Violet Jacobs and Mary Seacole. We compare the results obtained by the conditional model to their separate individually trained models and find that while the conditional model is a more challenging task, it has its benefits in collaborative writing specifically in the case of out-of-context generation.
Rahul is an MSc Artificial Intelligence student at the University of Aberdeen.
6. Michael McCall - Development of Social Media Clustering Algorithms for Capita
The project aims to design a social media text clustering algorithm for Capita to integrate into their wider Social Media and Research Toolkit (SMART). This would allow for faster analysis of incoming social media posts, event detection, and trend analysis. Much of the project focuses on natural language processing, taking social media posts and transforming them into a format suitable for machine learning.
Michael is an MSc Business Analytics student from RGU with a love for all things data – fascinated by its uses from sports to tackling the climate crisis.
The Aberdeen Meet-ups are hosted by Code The City and are supported by The Data Lab, Scotland IS and Opportunity North East.
Photo by Brook Cagle on Unsplash