We are a (still) young and growing institute with innovative projects and motivated young researchers. The focus of our research and teaching is on Artificial Intelligence and Machine Learning.
We develop and study computational models and algorithms that permit computers to perceive and 'understand' aspects of the external world, where we interpret 'perception' in the widest sense of the word, as the extraction of useful high-level information and knowledge from complex, possibly low-level data (audio, video, images, sensor data, texts, or even the Internet).
This requires fundamental and applied research on AI, machine learning, pattern recognition, text / data / social media mining, signal processing, statistical data modeling and classification.
Our current research has a particular focus on intelligent audio and music processing. Further research topics include biometrics, cryptography, personalization, and recommender systems.
Our goal is to offer state-of-the-art research and teaching in this area, and to provide a teaching environment that permits students to get involved in real research projects as early as possible.
The work of CP researchers on gender bias in neural Information Retrieval models in news
Decisions made by learning-based systems can be discriminatory or racist, caused by the underlying biases in data and models. Navid Rekab-Saz and Markus Schedl from the Computational Perception Institute investigate this topic in the context of search engines, and find out that use deep learning models are particularly gender biased. Read the full story here:
CP researchers present their latest work on Trustworthy Machine Learning
Despite the successes of Deep Neural Networks in many domains, it has been shown that they are very brittle when encountered with adversarial examples, that are instances with small, intentional perturbations causing these models to make false predictions.
In their latest work, CP researchers Hamid Eghbal-zadeh, Khaled Koutini, Paul Primus, Verena Haunschmid, and Gerhard Widmer look for the causes of this brittleness, by studying the common inductive biases used in Deep Learning.
They investigate Data Augmentation, that is a widely-used technique in deep learning for extending the training data by using inductive biases and domain expertise.
Their study reveals that although these methods have been proposed to improve performance, they can result in severe adversarial vulnerabilities.
Read their full story here:
click here to view the paper
CP Researchers attend the "Microsoft Research Frontiers in Machine Learning 2020" virtual conference
CP researcher Hamid Eghbal-zadeh, and CP PhD students Khaled Koutini, Florian Henkel, Paul Primus, and Shreyan Chowdhury have been invited to attend the "Research Frontiers in Machine Learning 2020" virtual conference organized by Microsoft Research.
During this four-day virtual event, researchers and academics from around the world presented their latest work, and discussed in live sessions with the attendees; on the theory and practice of generalisation in machine learning, trustworthy in deep learning, algorithmic fairness, causal machine learning, and interpretable machine learning.
Find more information about this event, and watch the recorded videos here:
Top Ranks for CP Teams in the DCASE 2020 Challenge
CP Team members Khaled Koutini, Florian Henkel, Paul Primus, Verena Haunschmid, Hamid Eghbal-zadeh, and Gerhard Widmer achieve top ranks in the DCASE 2020 Acoustic Event Detection and Classification Challenge!