You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: index.html
+27-9
Original file line number
Diff line number
Diff line change
@@ -68,13 +68,13 @@
68
68
I am a <istyle="color:red;">Research MSc student</i> at <ahref="https://www.polymtl.ca/" target="_blank">Polytechnique Montréal</a> supervised by Associate Professor<ahref="https://www.polymtl.ca/expertises/cappart-quentin-0/" target="_blank"> Quentin Cappart</a> and Full Professor <ahref="https://www.polymtl.ca/expertises/rousseau-louis-martin/" target="_blank">Louis-Martin Rousseau</a>.
69
69
</p>
70
70
<p>
71
-
To quickly introduce myself, I was admitted to <ahref="https://www.polytechnique.edu/" target="_blank">Polytechnique</a> (France) in 2018, I did a bachelor’s degree in applied mathematics/Computer Science. After a research internship at Polytechnique Montreal in the summer of 2021, I decided to continue in the Master Thesis program in the <ahref="https://corail-research.github.io/" target="_blank"> Corail Research Group</a>.
71
+
To quickly introduce myself, I come from <ahref="#maps" target="_self">Reunion island</a>, I was admitted to <ahref="https://www.polytechnique.edu/" target="_blank">Polytechnique</a> (France) in 2018, I did a bachelor’s degree in applied mathematics/Computer Science. After a research internship at Polytechnique Montreal in the summer of 2021, I decided to continue in the Master Thesis program in the <ahref="https://corail-research.github.io/" target="_blank"> Corail Research Group</a>.
72
72
<br>
73
-
My research interest lies at the crossroad of Operational Research and Deep Learning.
73
+
My research interest lies at the crossroad of Operational Research and Deep Learning, with a focus on :
74
74
<ul>
75
75
<li>Constraint Programming</li>
76
76
<li>Graph Representation Learning</li>
77
-
<li>Reinforcement Learning</li>
77
+
<li>Deep Reinforcement Learning</li>
78
78
</ul>
79
79
Corail's research area is Operations Research (OR). I joined the open-source project <ahref="https://github.com/corail-research/SeaPearl.jl" target="_blank">SeaPearl</a> and am now its main contributor. The goal is to develop an intelligent solver guided by RL to solve NP-hard discrete optimization problems using constraint programming (CP).
80
80
</p>
@@ -175,6 +175,9 @@ <h2>News</h2>
175
175
<br>
176
176
I am an active lead maintainer of the project SeaPearl, an intelligent solver guided by RL to solve NP-hard discrete optimization problems using constraint programming.
177
177
<br>
178
+
<br>
179
+
Click <ahref="#seapearl" target="_self">here</a> for detailed explanations.
This project was carried out in the framework of the IFT6167 seminar led by Prof. <aclass="tog" href="https://mila.quebec/personne/irina-rish/" target="_blank">Irina Rish</a> (Mila, Québec).
196
199
<br>
197
200
<br>
198
-
In this paper, we aim to show, regarding <aclass="tog" href="https://en.wikipedia.org/wiki/Sentiment_analysis" target="_blank">Natural Language Sentiment Analysis</a>, that there exists a relationship between model size and robustness to adversarial attacks. Ultimately, uncovering the emergence of power laws and testing the robustness of language model with scale. We evaluate the performance of various <aclass="tog" href="https://github.com/EleutherAI/" target="_blank">Eleuther AI</a> GPT models such as GPT-Neo 125M, GPT-Neo 1.3B, GPT-Neo 2.7B, GPT-J 6B against adversarial attacks. We fined-tuned (trained on adversarial example) our different GPT models on common datasets (Rotten Tomatoes, IMDB...) and evaluated them separately to quantify the effects of scale on adversarial training.
201
+
In this work, we aim to show, regarding <aclass="tog" href="https://en.wikipedia.org/wiki/Sentiment_analysis" target="_blank">Natural Language Sentiment Analysis</a>, that there exists a relationship between model size and robustness to adversarial attacks. Ultimately, uncovering the emergence of power laws and testing the robustness of language model with scale. We evaluate the performance of various <aclass="tog" href="https://github.com/EleutherAI/" target="_blank">Eleuther AI</a> GPT models such as GPT-Neo 125M, GPT-Neo 1.3B, GPT-Neo 2.7B, GPT-J 6B against adversarial attacks. We fined-tuned (trained on adversarial example) our different GPT models on common datasets (Rotten Tomatoes, IMDB...) and evaluated them separately to quantify the effects of scale on adversarial training.
In collaboration with Sariah Al Saati, Mehdi Benharrats, Swann Chelly and Pierre Tessier, this report proposes a method for the coverage of a rescue zone with a swarm of UAV’s in order to detect possible target of interest. The method is based on <aclass="tog" href="https://ieeexplore.ieee.org/document/8963663" target="_blank">Collaborative Reinforcement Learning</a>. It also presents a pipeline to locate points of interest in 3D from a set of 2D images using <aclass="tog" href="https://towardsdatascience.com/inverse-projection-transformation-c866ccedef1c" target="_blank">Inverse Projection Transformation</a> and 3D ray clustering.
236
239
<br>
237
240
<br>
238
241
<pstyle="padding:0%;text-align:center">
@@ -293,7 +296,7 @@ <h2>News</h2>
293
296
Mar' 20 – June' 20
294
297
<br>
295
298
<br>
296
-
In collaboration with Elsa Deville <imgsrc="images/whale.png" width="18px"> (MS, Imperial College), the objective of our project is to render a realistic 3D marine scene using only randomized <aclass="tog" href="https://en.wikipedia.org/wiki/Procedural_modeling" target="_blank">procedural modeling</a> (OpenGL).
299
+
In collaboration with Elsa Deville <imgsrc="images/whale.png" width="18px"> (MS, Imperial College), the objective of our project is to fully render a realistic 3D marine scene using only randomized <aclass="tog" href="https://en.wikipedia.org/wiki/Procedural_modeling" target="_blank">procedural modeling</a> (OpenGL).
297
300
<br>
298
301
<br>
299
302
The project implements different visual elements among them terrain generation using <aclass="tog" href="https://www.scratchapixel.com/lessons/procedural-generation-virtual-worlds/perlin-noise-part-2/perlin-noise-terrain-mesh" target="_blank">Perlin Noise</a>, a realistic simulation of a fish swarm movement based on <aclass="tog" href="https://www.cs.toronto.edu/~dt/siggraph97-course/cwr87/" target="_blank">this paper</a> and realistic-looking ocean waves. (<aclass="tog" href=" https://en.wikipedia.org/wiki/Trochoidal_wave" target="_blank">Trochoidal Waves</a> : exact solution of the Euler equations for periodic surface gravity waves).
<tdstyle="padding:20px;width:75%" valign="center" align="left ">This is where I come from ! <imgsrc="images/reunion.png" width="18px"><imgsrc="images/reunion.png" width="18px"><imgsrc="images/reunion.png" width="18px"></td>
0 commit comments