Research Training Workshop on NeuroAI
Title: Unsupervised discovery of sequential patterns in continuous temporal data using VQ-VAEs
Description: In this workshop, we will be covering theory and practice for a relatively recent technique using vector-quantized autoencoders (VQVAE) for multidimensional temporal data clustering using deep neural networks. We used VQVAEs to learn a motion vocabulary for temporal data (e.g. a human skeleton-based motion capture dataset), and we suggest how it might be extended to EEG data or similar. Attendees will learn the theory behind autoencoders and their vector-quantized variants, used extensively for recent deep neural multimodal tokenization (e.g. for speech, video, images), and come away with understanding of Python code of VQVAE applied to human neural data.
Facilitators:
Dr. Angelica Lim is an associate professor in the School of Computing Science at Simon Fraser University (SFU) and Director of the ROSIE Lab (Robots with Social Intelligence and Empathy). She develops artificial intelligence models of nonverbal communication, including facial expressions, body gestures and speech prosody, to build empathic, context-aware and compassionate machines. Her work is grounded in interdisciplinary collaboration with psychologists, cognitive scientists, clinicians and neuroscientists, advancing responsible and human-centered robotics. She is the creator of the SFU CS Teaching Toolkit, the author of Python Practice Lab (Princeton Univ. Press), and has received multiple teaching awards at the university level.
Payam Jome Yazdian is a Ph.D. student in the School of Computing Science at Simon Fraser University. He received his BSc degree in Software Engineering from the University of Sistan & Baluchestan, Iran, in 2014 and his MSc in Artificial Intelligence and Robotics from the University of Tehran, Iran in 2017. His main research interests include Affective Computing, Human-Robot Interaction (HRI), Cognitive Science, and Machine Learning.
Requirements/Prerequisites:
- A laptop with a web browser capable of running Google Colab / Python Jupyter notebooks
- Basic knowledge of Python or other programming knowledge
- Experience with data processing
Date/Time: Thursday March 19th, 10:30am-12pm PDT
Location: Burnaby campus & online
Register by emailing inn@sfu.ca. Please indicate whether you will attend in-person or via Zoom. This event is open to all but space is limited and priority will be given to INN trainee members.
Register Now
Neuroscience and Neurotechnology Seminar
Title: The good of skill: Skill expression and affective cognitive states
Abstract: Within the cognitive science and philosophy of skill, there is a widely held view that certain mental (i.e., neurocognitive and phenomenological) states are necessary for successful expert-level skill expression. For example, one might argue that being an expert skier involves behavioral reflexes which rely on the neural basis for automaticity. This assumption has been used to support the idea that there is something good about skill, insofar as it plays a positive role in human well-being: if automaticity is considered a positive affective cognitive state, and if skill expression elicits such a cognitive state, then this is one important basis for thinking that there is something good about skill. In this talk, I argue that the assumption that successful skill expression depends on specific mental states might be empirically ill-informed. On this basis, I argue that the affective dimension of cognitive states similarly comes apart from successful skill expression. This opens new avenues for considering the relationship between skill, the brain, and the good of skill.
Speaker: Dr. Zara Anwarzai, Assistant Professor, Department of Philosophy, SFU
Date/Time: Thursday March 26, 2:30pm
In-person Location: Big Data Hub Presentation Studio, ASB 10900, SFU Burnaby Campus
Online: https://sfu.zoom.us/j/84906133107?pwd=ASwc2prprtyUmBwBW5ZUGb8JRR4KaE.1
