Multidisciplinary Institute in Artificial intelligence

DESCRIPTION

Our research team, lead by Martial Mermillod and Sophie Achard is developping bio-inspired algorithms to improve the behaviour of deep neural networks (DNN).

Deep Neural Networks (DNN) have recently shown important limitations for different computational tasks. Our aim is (1) to use innovative mathematical algorithms to understand how DNN solve these computational problems and (2) to provide new architectures to address these limitations in a more efficient manner than the state-of-the-art.

ABOUT OUR RESEARCH

RESEARCHING BIO-INSPIRED WAYS TO IMPROVE THE ROBUSTNESS OF CONVOLUTIONAL NEURAL NETWORKS (CNN) TO ADVERSARIAL PERTURBATIONS

CNN suffer from adversarial perturbations (Goodfellow et al., 2015), i.e. small shifts in input images which trick the CNN into misclassifying the image. One explanation of this problem could be found in the oversensitiy of these networks to texture information (Geirhos et al., 2019), and more generally, because the features these networks use to classify an image seem to strongly differ from humans.

Recent studies seem indeed to point out that CNN use a wide range of different features to classify images, many of them being irrelevant to humans, and possibly great targets to craft adversarial perturbations. (Ilyas et al., 2019). For example, high spatial frequencies (which include texture information) are vastly used by naturally trained CNN, to the point where images filtered to a degree where only the very high frequencies remain still remain correctly categorized by state of the art CNN on datasets such as ImageNet. (Yin et al., 2019).

Why are CNN fooled by these perturbations, while human vision remains untouched ? Psychological and neuroscientific experiments have shown that human vision system relies heavily on spatial frequency processing of visual information, as well as recurrent, proactive processing (Bar, 2007) of the visual input, where higher level brain areas guide the lower level vision processing zones toward recognition of an object or a scene.

We currently focus on two main perspectives following these leads namely (1) developping algorithms able to mimick the recurrent, spatial filtering processes which take place in the human brain, and(2) finding way to highlight the "non-robust" features of a CNN which are the most vulnerable to adversarial perturbations.
 

SELECTED LIST OF PUBLICATIONS 

• Mainsant, M., Solinas, M., Reyboz, M., Godin, C., & Mermillod, M. Dream Net: a privacy preserving continual learning model for face emotion recognition. Proceedings of the 9th International Conference on Affective Computing & Intelligent Interaction (ACII 2021).
   
• Cohendet, R., Solinas, M., Bernhard, R., Reyboz, M., Moellic, P.A., Bourrier, Y., & Mermillod, M. (2021). Impact of reverberation through deep neural networks on adversarial perturbations. In 2021 20th IEEE International Conference on Machine Learning and Applications (ICMLA). IEEE.
   
• Bernhard ,R., Moellic, P.A., Mermillod, M., Bourrier, Y., Cohendet, R., Solinas, M., & Reyboz, M. (2021). Impact of Spatial Frequency Based Constraints on Adversarial Robustness. In 2021 IEEE International Joint Conference on Neural Networks (IJCNN).
   
• Solinas, M., Rousset, S., Cohendet, R., Bourrier, Y., Mainsant, M., Molnos, A., Reyboz, M. & Mermillod, M. (2021). Beneficial Effect of Combined Replay for Continual Learning.ICAART 2020 - Proceedings of the 13th International Conference on Agents and Artificial Intelligence, Volume 2. 205-217.
   
• Solinas, M., Galiez, C., Cohendet, R., Rousset, S., Reyboz, M., & Mermillod, M. (2020). A universal property of autoencoders and application to transfer learning. In 2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA) (pp. 877-882). IEEE.
   
• Lacroix, A.,Nalborczyk, L., Dutheil, F., Kovarski, K., Chokron, S., Garrido, M., Gomot, M., & Mermillod, M. (2021). High spatial frequency filtered primes hastens happy faces categorization in autistic adults. Brain and Cognition, 155, 105811.
   
• Entzmann, L., Guyader, N., Kauffmann, L., Lenouvel, J., Charles, C., Peyrin, C., ... & Mermillod, M. (2021). The Role of Emotional Content and Perceptual Saliency During the Programming of Saccades Toward Faces. Cognitive Science, 45(10), e13042.
   
• Park, G., Kim, H., Mermillod, M., & Thayer, J. F. (2021). The Modulation of Cardiac Vagal Tone on Attentional Orienting of Fair-Related Faces: Low HRV is Associated with Faster Attentional Engagement to Fair-Relevant Stimuli. Cognitive, Affective, & Behavioral Neuroscience, 1-15.
   
• Bret, A., Beffara, B., Mierop, A., & Mermillod, M. (2021). Differentiated evaluation of counter-conditioned stimuli as a function of right-wing authoritarianism. Social Psychological Bulletin, 16(2), 1-26.
   
• Mermillod, M., & Morisseau, T. (2021). Protect Others to Protect Myself: A Weakness of Western Countries in the face of Current and Future Pandemics? Psychological and Neuroscientific Perspectives. Frontiers in Integrative Neuroscience, 15, 8.
   
• Serole, C., Auclair, C., Prunet, D., Charkhabi, M., Lesage, F.X., Baker, J.S., Mermillod, M., Gerbaud, L., Dutheil, F. (2021). The Forgotten Health-Care Occupations at Risk of Burnout – A Burnout, Job Demand-Control-Support, and Effort-Reward Imbalance Survey. Journal of Occupational and Environmental Medicine, 63 (7), e416-e425.
   
• Dutheil, F., Comptour, A., Morlon, R., Mermillod, M., Pereira, B., Baker, J. S., Charkhabi, M., Clinchamps, M., & Bourdel, N. (2021). Autism spectrum disorder and air pollution: a systematic review and meta-analysis. Environmental Pollution, 116856.
   
• Merlhiot, G., Mondillon, L., Méot, A., Dutheil, F., & Mermillod, M. (2021). Facial width-to-height ratio underlies perceived dominance on facial emotional expressions. Personality and Individual Differences, 172, 110583.
   
• Shankland, R., Favre, P., Kotsou, I., & Mermillod, M. (2021). Mindfulness and De-automatization: Effect of Mindfulness-Based Interventions on Emotional Facial Expressions Processing. Mindfulness, 12(1), 226-239.
   
• Clinchamps, M., Auclair, C., Prunet, D., Pfabigan, D., Lesage, F. X., Baker, J. S., Parreira, L., Mermillod, M., Gerbaud, L., & Dutheil, F. (2021). Burnout Among Hospital Non-Healthcare Staff: Influence of Job-Demand-Control Support, and Effort-Reward Imbalance. Journal of Occupational and Environmental Medicine, 63(1), e13-e20.