Let me introduce myself

Hi there! My name is Thymios!

I am a fourth year PhD candidate in the Computer Science (CS) department at the University of Illinois Urbana-Champaign (UIUC) where I investigate machine learning for audio processing problems with my advisor Prof. Paris Smaragdis. I was also fortunate enough to work as a research intern and a student researcher at Google AI, under the supervision of Dr. Scott Wisdom and Dr. John Hershey, and as a research intern at Facebook Reality Labs (FRL), where I was managed by Dr. Anurag Kumar. My current research focuses on how we can utilize neural networks towards more generalizable and robust audio source separation. I am particularly interested in unsupervised ways of learning directly from mixtures of audio signals as well as making source separation neural architectures accessible to everyone.
Before that, I graduated with a diploma (Bachelor and MEng equivalent) in Electrical and Computer Engineering (ECE) from the National Technical University of Athens (NTUA). In my thesis, I was investigating how nonlinear recurrence information from the phase space of phonemes and manifold learning could be utilized towards more efficient speech emotion recognition.
For more information, please check my academic CV.
Besides research, I love traveling, playing the guitar and playing soccer. During summer vacations, I spend most of my time playing beach rackets especially at the beautiful island of Santorini, where I also come from. If you also like etymology, Thymios derives from the Ancient Greek name “Euthýmios (Ευθύμιος)”, composed of two elements: “eû ‎(εὖ)” (well) plus “thūmós (θῡμός)” (soul, as the seat of emotion, feeling, life, desire, will, temper, passion), meaning “in good spirits, of good cheer, clear”.
What's up?

News Feed

Paper Accepted! 06/20/2021 Unsupervised federated learning for speech enhancement is going to be presented at WASPAA 2021! Check our cool paper, "Separate but Together: Unsupervised Federated Learning for Speech Enhancement from non-IID Data." here: [pdf] [code]
Status Change 05/17/2021 I am going to spend my summer as a research intern for Facebook Reality Labs (FRL) where I will be supervised by Dr. Anurag Kumar !
Status Change 05/07/2021 After a long and fruitful collaboration with the sound separation team at Google AI Perception, it's time to say goodbye! I would like to express my sincere gratitude towards Scott Wisdom and John Hershey who were not simply my managers but also acted as my advisors and mentors! We did great things together, I will miss you guys!
Presentation 05/02/2021 I am giving a virtual presention of our paper "Into the Wild with AudioScope: Unsupervised Audio-Visual Separation of On-Screen Sounds." at ICLR 2021! You can check the video-presentation here! [paper] [slides].
Honors & Awards 04/28/2021 I was chosen as a finalist (top 3.5% of the applications) for the renowned "Facebook PhD Fellowship".
Paper Accepted! 01/28/2021 Our paper with Dimitris "Unified Gradient Reweighting for Model Biasing with Applications to Source Separation" has been accepted at ICASSP 2021! Please take a look at how you can exploit biases in NNs here: [pdf] [code]
Paper Accepted! 01/12/2021 AudioScope has been accepted at ICLR 2021! You can take a look at our paper: "Into the Wild with AudioScope: Unsupervised Audio-Visual Separation of On-Screen Sounds" [pdf]
Presentation 12/11/2020 I am giving a virtual oral presention of our paper "Self-Supervised Audio-Visual Separation of On-Screen Sounds from Unlabeled Videos" for SAS workshop at NeurIPS 2020! You can check the video-presentation here! [workshop] [video].
Honors & Awards 09/26/2020 I was selected to represent UIUC in the national competition for the renowned Apple PhD fellowship!
Paper Accepted! 09/25/2020 MixIT has been accepted as a Spotlight paper at NeurIPS 2020! You can take a look at the paper here: "Unsupervised Sound Separation using Mixtures of Mixtures" [pdf]
Presentation 08/23/2020 I am giving a virtual oral presention of our paper "Sudo rm -rf: Efficient Networks for Universal Audio Source Separation" for Lecture Session 7 at MLSP 2020! You can check the video-presentation here! [session] [video] [slides] [pdf].
Paper Accepted! 06/30/2020 Sudo rm -rf has been accepted for an oral presentation at MLSP 2020! You can take a look at the paper and the code here: "Sudo rm -rf: Efficient Networks for Universal Audio Source Separation" [pdf] [code].
Paper Accepted! 06/30/2020 Our paper on unsupervised speech separation using mixtures of mixtures has been accepted at ICML 2020 workshop for self-supervision in audio and speech! You can take a look at the paper here: "Unsupervised Speech Separation Using Mixtures of Mixtures" [pdf]
Status Change 05/18/2020 I am going to spend my summer working at Google AI Perception!
Presentation 05/15/2019 I am giving a virtual oral presention of our paper "Two-Step Sound Source Separation: Training on Learned Latent Targets" for the session AUD-L2 of ICASSP 2020! [session] [presentation] [paper-ieee].
Presentation 05/15/2019 I am giving a virtual oral presention of our paper "Improving Universal Sound Separation Using Sound Classification" for the session AUD-L4 of ICASSP 2020! [session] [presentation] [paper-ieee].
Education Timeline

Education

PhD in Computer Science (CS) at the
University of Illinois at Urbana-Champaign (UIUC)
Urbana-Champaign, Illinois, USA | Aug 2018 - Current

  • Area: Artificial Intelligence
  • Current GPA: 4.00/4.00
  • Selected Courses: Computational Inference and Learning, Machine Learning for Signal Processing, Optimization in Computer Vision, Machine Learning, Deep Learning, Deep Learning Theory, Probabilistic and Approximate Computation
  • Advisor: Prof. Paris Smaragdis

Diploma (BS + Meng) in Electrical and Computer Engineering (ECE) at the
National Technical University of Athens (NTUA)
Athens, Greece | Oct 2012 - Jun 2018

  • Area: Computer Science
  • Highest Honors - top 2%
  • GPA: Cumulative 9.36/10.00 | Major 9.56/10.00
  • Thesis: "Manifold Learning and Nonlinear Recurrence Dynamics for Speech Emotion Recognition on Various Timescales" [pdf] [slides]
  • Advisor: Prof. Alexandros Potamianos

For more information, please check my academic CV.

Professional Timeline

Work Experience

Research Intern at Facebook Reality Labs (FRL)
Redmond, Washington, USA | May 2021 - Aug 2021

Self-supervised audio source separation at scale.

Student Researcher at Google AI
Urbana, Illinois, USA | Aug 2020 - May 2021

In-the-wild audio-visual universal sound source separation of on-screen sounds.

Graduate Teaching & Research Assistant at Audio Lab UIUC
Urbana-Champaign, Illinois, USA | Aug 2018 - Current

Fall 2018 - Spring 2020: Performing unsupervised audio source separation using deep clustering. Developing cost-efficient neural architectures for universal sound source separation.

Spring 2021: Teaching assistant for the course CS 446 Machine learning.
Fall 2020: Teaching assistant for the course CS 598 Machine learning for signal processing.

Research Intern at Google AI
Cambridge, Massachusetts, USA | May 2020 - Aug 2020

Unsupervised single-channel universal source separation using mixtures of mixtures. Purely unsupervised mixture invariant training obtains comparable results to fully supervised approaches!

Research Intern at Google AI
Cambridge, Massachusetts, USA | May 2019 - Aug 2019

Exploiting high-level semantic representations of sounds in order to boost the performance of universal sound source separation systems. The proposed architecture improved the state-of-the-art on sound separation when a variety of types of source signals might be active without the need of additional labels.

Machine Learning Engineer at Behavioral Signals
Los Angeles, California, USA | May 2017 - Jul 2018

Leading the machine learning infrastructure development. Building state-of-the-art models for speech emotion recognition and integrating them into product-level solutions. Automating the procedure of inference by implementing dynamic graph pipelines and optimizing feature extraction algorithms in Python.

Junior Researcher at ATHENA Research Center
Athens, Greece | May 2016 - Jul 2018

Building modelfs for real time speech emotion recognition and multimodal engagement detection (European project: BabyRobot).

Research Intern at SBA Research (IAESTE traineeship program)
Vienna, Austria | Jul 2016 - Aug 2016

End-2-end implementation (circuit level connections, BLE microprocessor programming, message encryption, Linux/Windows drivers and front-ends) of an automatic screen locker for increased computer security.

IT Advisory Intern at Ernst and Young
Athens, Greece | Jul 2015 - Oct 2015

Working at Piraeus Bank's database maintenance as an external partner. Performing financial data analysis and risk prediction.

Private Tutor
Athens, Greece | May 2013 - Jun 2017

Selected courses given: Maths, Physics, Differential Analysis, Algorithms.

For more information, please check my academic CV.

My contribution

Selected Research Papers

  • Tzinis, E., Wisdom, Remez, T., and Hershey, J. R., "Improving On-Screen Sound Separation for Open Domain Videos with Audio-Visual Self-Attention." [pdf] [video]
  • Tzinis, E., Casebeer, J., Wang, Z., and Smaragdis, P., "Separate but Together: Unsupervised Federated Learning for Speech Enhancement from non-IID Data." Submitted to IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), 2021. [pdf] [code]
  • Tzinis, E., Wisdom, S., Jensen, A., Hershey, S., Remez, T., Ellis, D. P., and Hershey, J. R., "Into the Wild with AudioScope: Unsupervised Audio-Visual Separation of On-Screen Sounds." To appear in International Conference on Learning Representations (ICLR), 2021. [DOI] [pdf] [website] [video] [slides].
  • Tzinis, E., Bralios, D., and Smaragdis, P., "Unified Gradient Reweighting for Model Biasing with Applications to Source Separation." To appear in International Conference in Acoustics, Speech and Signal Processing (ICASSP), 2021. [DOI] [pdf] [code] [video].
  • Wisdom, S., Tzinis, E., Erdogan, H., Weiss, R. J., Wilson, K., and Hershey, J. R., "Unsupervised Sound Separation Using Mixtures of Mixtures." In Advances in Neural Information Processing Systems (NeurIPS), 2020. [DOI] [pdf] [video].
  • Tzinis, E.,Wang, Z., and Smaragdis, P., "Sudo rm -rf: Efficient Networks for Universal Audio Source Separation." In IEEE International Workshop on Machine Learning for Signal Processing (MLSP), 2020. [DOI] [pdf] [code] [video].
  • Tzinis, E., Venkataramani, S., Wang, Z., Subakan, Y. C., and Smaragdis, P., "Two-Step Sound Source Separation: Training on Learned Latent Targets." In Proceedings of International Conference in Acoustics, Speech and Signal Processing (ICASSP), 2020. [DOI] [pdf] [code] [video].
  • Tzinis, E., Wisdom, S., Hershey, J.R., Jansen, A. and Ellis, D.P., "Improving Universal Sound Separation Using Sound Classification." In Proceedings of International Conference in Acoustics, Speech and Signal Processing (ICASSP), 2020 [DOI] [pdf] [video].
  • Tzinis, E., Venkataramani, S. and Smaragdis, P., "Unsupervised deep clustering for source separation: direct learning from mixtures using spatial information." In Proceedings of International Conference in Acoustics, Speech and Signal Processing (ICASSP), 2019, pp. 81-85. [DOI] [pdf] [code]


Thesis

    Tzinis, E., ""Manifold Learning and Nonlinear Recurrence Dynamics for Speech Emotion Recognition on Various Timescales." Diploma Thesis at the National Technical University of Athens, Electrical and Computer Engineering (ECE) Department. [pdf] [slides]

A full and always updated list of my research papers is also available at my scholar profile.

Get in Touch

Contact Me

Email me at: etzinis2@illinois.edu or etzinis@gmail.com
My office is located at the Siebel Center for Computer Science | No. 3332
201 N. Goodwin Ave. Urbana, IL, 61801, USA


“Γυμνοί ήλθομεν οι πάντες, γυμνοί και απελευσόμεθα.”
Αίσωπος, 620-560 π.Χ.
“We were all born naked and this is how we are going to die.”
Asopus, 620-560 BCE

Efthymios Tzinis © 2021

counter free