PhD student in Computer Science
Machine Learning & Language Understanding
University of Avignon - LIA
Who am I ?
I'm a young computer science enthusiast.
Always looking forward to improve state-of-the-art, and always looking to learn from the past.
About
Interested in computer science since my childhood, i just decided to follow the always learn and make more about computers way of life. Having the chance of being younger than the other persons around me, i always tried to use this advantage to improve my knowledge, doing personal and professional projects as much as i could.
Education
Just graduated from the "Ingénierie du Logiciel pour la Société Numérique" Master's degree, i decided to keep up with a thesis at the University of Avignon. During these 5 years of courses, we spent a lot of time learning computer and science basics, such as algorithmics, programming, optimization and other methods. We also dealed with project, process, and team management. This master was about software engineering, i have thus a strong background in programming through multiple languages (C++, JAVA, CUDA, Python, PHP, HTML, JS, C# ... ). In order to go further you can take a closer look at the full university program
A word about the thesis
Machine Learning (ML) techniques have allowed a great performance improvement of different challenging tasks (Spoken Language Understanding, Image Processing, Motion Management, Robotic ...). Among these methods, Neural Networks (NN), recently received a great interest from researchers due to their representation capability of complex internal structures in a low dimensional subspace. However, NNs employ document representations based on a basic level (because of their 1-Dimensionality). Therefore, these basic representations reveal little in way of document statistical structure by only considering each feature as a single entity, ignoring relations between them. My thesis propose to remedy this weakness by extending the hyper-complex numbers, also called quaternions, to neural networks called QNNs for Quaternion Neural Networks. Indeed, in a real-valued neural network, multidimensional features require to be reduced to a one dimensional vector before the learning process, while an appropriate solution is to process a multidimensional input as a single homogeneous entity. All these drawbacks can be alleviated thanks to the fourth dimensionality of quaternions and the hamilton product.
This thesis first proposes an integration of quaternions to state-of-the-art models such as : deep, recurrents, convolutionals, and adversarials neural networks. This integration rely on adapted and well-suited algorithms to fit the quaternion algebra specificities, and on a GPU implementation of quaternion neural networks. Then, a new type of neural network expressly designed for hyper-complex numbers wil be investigated, including a specific learning process and a well-adapted structure. This method has already reached promissing results that will be reported on my ResearchGate homepage and below.
Feel free to contact me if you want furthermore informations !
Publications
Highlights
- Speech Recognition with Quaternion Neural Networks
- Titouan Parcollet · Mirco Ravanelli · Mohamed Morchid · Georges Linarès · Renato De Mori
- NeurIPS 2018 (NIPS) - IRASL Workshop
- December 2-8 2018, Montréal, QC, (CA)
- The PyTorch-Kaldi Speech Recognition Toolkit
- Mirco Ravanelli · Titouan Parcollet · Yoshua Bengio
- ICASSP 2019
- May 12-17 2019, Brighton, (UK)
- Quaternion Recurrent Neural Networks
- Titouan Parcollet · Mirco Ravanelli · Mohamed Morchid · Georges Linarès · Chiheb Trabelsi · Renato De Mori · Yoshua Bengio
- ICLR 2019
- May 6-9 2019, New Orleans, (USA)
Awards and grants
- Quaternion Denoising Encoder-Decoder for theme identification of telephone conversations
- International Speech and Communication Association Grants - INTERSPEECH 2017
International conferences
- The PyTorch-Kaldi Speech Recognition Toolkit
- Mirco Ravanelli · Titouan Parcollet · Yoshua Bengio
- ICASSP 2019
- May 12-17 2019, Brighton, (UK)
- Bidirectional Quaternion Long-Short Term Memory Recurrent Neural Networks for Speech Recognition
- Titouan Parcollet · Mohamed Morchid · Georges Linarès · Renato De Mori
- ICASSP 2019
- May 12-17 2019, Brighton, (UK)
- Quaternion Convolutional Neural Networks for Heterogeneous Image Processing
- Titouan Parcollet · Mohamed Morchid · Georges Linarès
- ICASSP 2019
- May 12-17 2019, Brighton, (UK)
- Quaternion Recurrent Neural Networks
- Titouan Parcollet · Mirco Ravanelli · Mohamed Morchid · Georges Linarès · Chiheb Trabelsi · Renato De Mori · Yoshua Bengio
- ICLR 2019
- May 6-9 2019, New Orleans, (USA)
- Quaternion Convolutional Neural Networks for Theme Identification of Telephone Conversations
- Titouan Parcollet · Mohamed Morchid · Georges Linarès · Renato De Mori
- IEEE SLT 2018
- December 18-21 2018, Athens, (Greece)
- Speech Recognition with Quaternion Neural Networks
- Titouan Parcollet · Mirco Ravanelli · Mohamed Morchid · Georges Linarès · Renato De Mori
- NIPS 2018 - IRASL Workshop
- December 2-8 2018, Montréal, QC, (CA)
- Quaternion Convolutional Neural Networks for End-to-End Automatic Speech Recognition
- Titouan Parcollet · Ying Zhang · Mohamed Morchid · Chiheb Trabelsi · Georges Linarès · Renato De Mori · Yoshua Bengio
- INTERSPEECH 2018
- September 2-6 2018, Hyderabad, (India)
- Deep Quaternion Neural Networks for Spoken Language Understanding
- Titouan Parcollet · Mohamed Morchid · Georges Linarès
- ASRU 2017
- December 16-20 2017, Okinawa, (Japan)
- Quaternion Denoising Encoder-Decoder for theme identification of telephone conversations
- Titouan Parcollet · Mohamed Morchid · Georges Linarès
- Granted by ISCA
- INTERSPEECH 2017
- August 20-25 2017, Stockholm, (Sweden)
- Quaternion Neural Networks for Spoken Language Understanding
- Titouan Parcollet · Mohamed Morchid · Pierre-Michel Bousquet · Richard Dufour · Georges Linarès · Renato De Mori
- IEEE SLT 2016
- Decembre 13-16 2016, San Diego, (USA)
- Tracking Dialog States using an Author-Topic based Representation
- Richard Dufour · Mohamed Morchid · Titouan Parcollet
- IEEE SLT 2016
- Decembre 13-16 2016, San Diego, (USA)
National conferences
- Réseaux de neurones de quaternions pour le traitement du language
- Titouan Parcollet · Mohamed Morchid · Georges Linarès
- CORIA 2017
- March 29-31 2017, Marseille, (FRANCE)
Collaborations
ORKIS - PhD Fellowship
Orkis is a a french company based on Aix-en-provence, France. They are specialised in assets and data managements with well-known customers such as L'Oréal, Dassault, or the governement of France. The collaboration between the LIA (and thus the University of Avignon) started with a previous PhD fellowship which has now ended. In order to enhance their tools, they decided to pursue this collaboration with a new PhD student, me. My job involves many taks, from automatic indexation on speach recognition to image processing and machine learning. For more informations about the company, feel free to visit the Orkis website.
Teaching
All the tutorials and practical work classes are dispensed in addition to my standard research job
Year 2018/2019
| Teaching | Tutorials | Practical Work | Total |
| Tools for Machine Learning | 12h | 12h | |
| Total | 12h |
Year 2017/2018
| Teaching | Tutorials | Practical Work | Total |
| Advanced programming and projects (B.Sc.) | 12h | 19h | 31h |
| Total | 31h |
Year 2016/2017
| Teaching | Tutorials | Practical Work | Total |
| Advanced programming and projects (B.Sc.) | 12h | 18h | 30h |
| Agorithms and programming (B.Sc.) | 12h | 18h | 30h |
| Total | 60h |
Year 2015/2016
| Teaching | Tutorials | Practical Work | Total |
| C2I Certification (B.Sc.) | 60h | ||
| Total | 60h |
Year 2013/2014
| Teaching | Tutorials | Practical Work | Total |
| C2I Certification (B.Sc.) | 60h | ||
| Total | 60h |
Contact
Lab address
LIA - CERI
339 Chemin des Meinajariès
84911 Avignon, France
Mail & Phone
titouan.parcollet@alumni.univ-avignon.fr
+33 490 843 577