'Local Optima', the regular seminar of Release, the Department of Computer Science & Engineering society, was held for the fourth time at 1:30 p.m. on Saturday 4 January 2020 in AS-510. With more than 40 people attending, six talks were given.

Perhaps reflecting the society members' interest in machine learning, five of the six talks concerned machine learning. The first was a talk on attention methods by Kim Dong-min (the 19th cohort), covering the principles and applications of attention — which has drawn wide research interest since its application to neural machine translation in 2014 — and its cognitive significance as understood through its place within deep learning System 2.

▲ Speaker Kim Dong-min

Next came 'A Useful Introduction to Open Source' by Jung Yeon-jun (the 26th cohort), who graduated from the department with distinction and contributes extensively to Keras, the Python-based open-source machine learning library, followed by 'Recent Advances in Markov Chain Monte Carlo' by Kim Kyu-rae (the 22nd cohort), whose research on a novel scheduling approach based on Bayesian machine learning was accepted at MASCOTS 2019 (Qualis: A2).

▲ Speaker Jung Yeon-jun

There were also presentations on active learning, reinforcement learning, and Rust, the programming language drawing attention recently for its speed and memory safety.

▲ Speaker Kim Kyu-rae

Choi Kwang-hee (the 24th cohort), the former vice president of Release who has led the event, said with conviction: "I am glad that this event seems to have given many students the chance to look into a wider range of fields. I hope student-led academic events like this continue and develop further."

Generous refreshments and pizza were provided throughout the event, and a prize draw was also held.