Opportunities

Subject:

The vast majority of people who live in big cities face long days, between extended working hours and time spent in transport, it’s clear that a large proportion of them are constantly tired [1].

Yet there are days when you feel less tired, even if you’ve completed several tasks, even if the metro has been disrupted, even if you’ve spent more hours than usual at the office…

This feeling of fatigue may be partly due to the person’s lifestyle (lack of sleep, poor diet…). But part of the fatigue may also be linked to the subjects’ cognitive conditions. Academic literature shows that the sympathetic system can influence daily fatigue [2]. The sympathetic system is responsible for fight-or-flight reactions, and certain emotions such as stress and anxiety are closely linked to this system.

How can the sympathetic system influence subject fatigue? This is the general question that this course will attempt to answer.

More specifically, the trainee will :

• Examine the academic literature to better understand the subject and find protocols for collecting data on stress, anxiety and fatigue.
• Collect biometric data related to these cognitive states.
• Process and analyze the data collected.
• Develop machine learning algorithms capable of identifying each situation individually.
• Identify the influence of stress and anxiety on daily fatigue.
• Write a scientific report.

Comments :

• This is a multi-disciplinary internship linked to the themes of human cognition and machine learning algorithms.
• The trainee must be motivated to work in these two areas, be organized and have good written and oral communication skills.
• The final objective is to write an article to be submitted to a scientific journal.

Practical information:

• The internship will pay 1,200 euros gross per month.
• The internship will last 6 months and is scheduled to start in February 2024.

References:

[1] Madden, D. (2022). Tired city: on the politics of urban exhaustion. City, 26(4), 559-561.

[2] Tanaka, M., Mizuno, K., Yamaguti, K. et al. Autonomic nervous alterations associated with daily level of fatigue. Behav Brain

Contact details:

Aakash SONI – aakash.soni@ece.fr

Guilherme MEDEIROS MACHADO – gmedeirosmachado@ece.fr

Working environment :

In recent years, language models such as BERT, GPT and LLM have considerably transformed the field of automatic language processing (ALP). Their ability to capture complex contextual information has led to outstanding performance on a variety of tasks and in multiple languages. These models have set new standards in natural language understanding and text generation.

In the context of this internship, we are focusing specifically on multilingual sentiment analysis. Sentiment analysis is a fundamental task in natural language processing (NLP), involving the evaluation of sentiments and the tonality expressed (polarity) in a text [1,2]. Using these AI models, we are developing linguistically powerful sentiment analysis systems, capable of efficiently processing texts from a variety of sources and in different languages. By integrating these models into our approach, we improve the accuracy of the model and the versatility of sentiment analysis on a global scale.

Problem :

Despite significant advances in the GPT and Transformers models, understanding cultural and linguistic nuances in multilingual sentiment analysis remains a challenge. How can we adapt these models to better capture the diversity of emotional expressions across different languages and cultures, and what strategies can be implemented to optimize their performance in this context?

Tasks and objectives :

• Multilingual data extraction: Collect varied and representative datasets, covering different languages and sources, to train and evaluate our models.
• Data pre-processing: Clean and pre-process text data to make it ready for analysis.
• Sentiment analysis: Implement machine learning models adapted to sentiment analysis in a multilingual context, exploring architectures such as Transformers, LLMs and GPT generative models.
• Information extraction: Analyze sentiment behavior in different languages.

Profile :

• In-depth exploration of deep learning methods.
• Extensive experience of the Python programming language.
• Interest in data processing and pre-trained models on different languages.
• Interest in reading and writing scientific articles, as well as curiosity about research challenges.

Practical information:

• The internship will pay 1,200 euros gross per month.
• The internship will last 6 months and is scheduled to start in February 2024.

References:

[1] Deshpande, Ameet Shridhar et al. “When is BERT Multilingual? Isolating Crucial Ingredients for Cross-lingual Transfer.” North American Chapter of the Association for Computational Linguistics (2021).

[2] Dou Hu, Lingwei Wei, Yaxin Liu, Wei Zhou, and Songlin Hu. 2023. UCAS-IIE-NLP at SemEval-2023 Task 12: Enhancing Generalization of Multilingual BERT for Low-resource Sentiment Analysis. In Proceedings of the 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 1849-1857, Toronto, Canada. Association for Computational Linguistics.

Contact person:

Assia SOUKANE- assia.soukane@ece.fr

Faiza BELBACHIR – fbelbachir@ece.fr

Subject:

Cancer patients may have to limit their food intake for a number of reasons. Firstly, cancer pathology in itself can induce anorexia (1). Secondly, certain tumor locations cause symptoms that restrict eating: dysphagia in ENT and esophageal tumors, abdominal pain and vomiting in gastric and pancreatic tumors (2). Finally, certain anti-cancer treatments such as systemic chemotherapy, targeted therapy and radiotherapy can be directly responsible for symptoms that reduce patient intake: mucositis, taste disorders, nausea, vomiting, dysphagia, anorexia, transit disorders (3). This can lead to weight loss and even the development of undernutrition (4). Alterations in nutritional status can lead to micro-nutritional deficiencies (5).

Micro-nutritional deficiencies are known to have specific clinico-biological repercussions. For example, deficiencies in vitamin B9, vitamin B12, iron and copper can lead to anemia (6, 7, 8, 9). Zinc, selenium, iron and vitamin D deficiencies are associated with immune dysfunction. Vitamin B1 and selenium deficiencies can weaken heart function. Vitamin PP and zinc deficiencies can lead to skin dermatoses and problems with skin healing. Deficiencies in vitamins B1, B6, B12, PP and copper can lead to neurological complications.

These clinico-biological consequences are problematic for cancer patients, already weakened by their pathology and its specific treatments. These are immunocompromised patients, particularly at risk of anemia due to bone marrow toxicity, under certain chemotherapies. Some anti-cancer therapies also have cardiac, neurological and dermatological toxicities.

It is therefore of great interest to be able to detect and correct any micro-nutritional deficiencies present in patients undergoing cancer treatment.

However, few studies have analyzed the prevalence of micro-nutritional deficiencies in oncology and their potential clinico-biological consequences, apart from vitamin D and iron.

In addition, the detection of these deficiencies requires blood sampling, the analysis of which for certain micronutrients (zinc, selenium, copper, vitamins B1, B6, PP) is outsourced to specialized external laboratories. This poses the problems of tube damage during transport, long delays in obtaining results and significant financial costs.

The development of predictive algorithms, using risk scores for micronutrient deficiencies, could be of interest in targeting patients in whom blood sampling for plasma micronutrient assays would be of interest. We could then consider performing blood tests only on patients with a high-risk deficit score.

The project is a collaboration between the Institut Gustave Roussy (IGR) and L’ECE – Paris. The team is made up of 2 IGR doctors specialized in clinical nutrition for cancer patients and ECE research professors.

Tasks and objectives :

• To determine the prevalence of micro-nutritional deficits in a cohort of adult cancer patients hospitalized in an oncology Nutrition-Suite Care and Rehabilitation department.
• Identify potential factors associated (clinical, paraclinical, anamnestic) with micro-nutritional deficits in the patient population studied.
• Establish predictive risk scores for micro-nutritional deficiencies in this patient population studied.

Practical information:

• The internship will pay 1,200 euros gross per month.
• The internship will last 6 months and is scheduled to start in February 2024.
• Part of this work will be carried out at the IGR.

References:

[1] Hariyanto TI, Kurniawan A. Appetite problem in cancer patients: Pathophysiology, diagnosis, and treatment. Cancer Treat Res Commun. 2021;27:100336

[2] Navari RM. Nausea and Vomiting in Advanced Cancer. Curr Treat Options Oncol. 2020 Feb 5;21(2):14.

[3] Gupta K, Walton R, Kataria SP. Chemotherapy-Induced Nausea and Vomiting: Pathogenesis, Recommendations, and New Trends. Cancer Treat Res Commun. 2021;26:100278

[4] Bossi P, Delrio P, Mascheroni A, Zanetti M. The Spectrum of Malnutrition/Cachexia/Sarcopenia in Oncology According to Different Cancer Types and Settings: A Narrative Review. Nutrients. 2021 Jun 9;13(6):1980

[5] Clement DS, Tesselaar ME, van Leerdam ME, Srirajaskanthan R, Ramage JK. Nutritional and vitamin status in patients with neuroendocrine neoplasms World J Gastroenterol. 2019 Mar 14;25(10):1171-1184

[6] Lazarchick J. Update on anemia and neutropenia in copper deficiency. Curr Opin Hematol. 2012 Jan;19(1):58-60

[7] Green R. Vitamin B12 deficiency from the perspective of a practicing hematologist. Blood. 2017 May 11;129(19):2603-2611

[8] Tardy AL, Pouteau E, Marquez D, Yilmaz C, Scholey A. Vitamins and Minerals for Energy, Fatigue and Cognition: A Narrative Review of the Biochemical and Clinical Evidence. Nutrients. 2020 Jan 16;12(1):228

[9]Pasricha SR, Tye-Din J, Muckenthaler MU, Swinkels DW. Iron deficiency. Lancet. 2021 Jan 16;397(10270):233-248

Contact details:

Amine JAOUADI – ajaouadi@ece.fr

Guilherme MEDEIROS MACHADO – gmedeirosmachado@ece.fr

Subject:

This internship falls within the scope of the implementation of 6G network technologies for energy management of green renewable sources within the framework of smart grids. The aim of this project is to propose an algorithm based on machine learning, to help decision-makers in the demand modulation program. Based on this algorithm, the consumer adopts the most suitable program to suit his needs and reduce his electricity bill.

The aim of DR demand modulation programs is to modulate the electricity load curve by shifting loads, smoothing peaks or filling troughs.

• Load-shifting consists in shifting the demand of an electrical device, i.e. postponing or bringing forward a demand from one time slot to another.
• Reducing peak demand, or peak clipping, can be achieved by reducing or interrupting a specific electrical use [1].
• While the last two Demand Response action strategies seek to flatten the load curve by smoothing out peaks in demand, valley filling increases load during periods of lower demand.

In this project, we advocate artificial intelligence via machine learning as a basic tool to help the consumer adopt a demand modulation program. Artificial intelligence has proved its worth in controlling energy production and consumption, as well as energy supply and demand [2-5]. More specifically, machine learning algorithms can be used to finely assess supply and demand, and make highly accurate forecasts [6,7].

NGSG smart grids enable the collection of consumption data that requires continuous monitoring, analysis and interpretation. In turn, wind and solar farm operators also collect information on the quantity and energy content of renewable sources exported to the grid [8], to ensure that electricity supply meets demand. In this context, the use of a machine learning algorithm will have a decisive impact on energy production and contribute to optimizing energy consumption.

The methodology required to complete the work :

1. Read the academic literature to gain a better understanding of the subject.
2. Collect energy consumption data.
3. Process and analyze the data collected. 4- Develop machine learning algorithms capable of proposing solutions in line with DR programs, taking into account user needs and power grid capacity.

Comments :

• The trainee must be motivated to work in the field of machine learning, be organized and have good written and oral communication skills.
• The final objective is to write an article to be submitted to a scientific journal.

Practical information:

• The internship will pay 1,200 euros gross per month.
• The internship will last 6 months and is scheduled to start in February 2024.

References:

[1] ADEME. Report on electricity load shedding in France. Assessment of the potential for load shedding through process modulation in industry and the service sector in mainland France, 2017

[2] Alreshidi, E. (2019). “Smart Sustainable Agriculture (SSA) Solution Underpinned by Internet of Things (IoT) and Artificial Intelligence (AI)”. Ijacsa 10, 03106. doi:10.14569/IJACSA.2019.0100513

[3] Pinto et al. 2019; Pinto, T., Morais, H., and Corchado, J. M. (2019). “Adaptive Entropy-Based Learning with Dynamic Artificial Neural Network. Neurocomputing”, 432-440. doi:10.1016/j.neucom.2018.09.092

[4] Ji et al, 2020; Ji, H., Alfarraj, O., and Tolba, A. (2020). “Artificial Intelligence-Empowered Edge of Vehicles: Architecture, Enabling Technologies, and Applications”. IEEE Access 8, 61020-61034. doi:10.1109/ACCESS.2020.2983609

[5] R.Naja, A.Soni, C. Carletti, “Optimal Energy Management for Electric Vehicles In V2G 6G-based Smart Grid Networks”, In Journal MDPI JSAN, To appear

[6] Ma and Zhai, 2019; Ma, Y.-J., and Zhai, M.-Y. (2019). “Day-Ahead Prediction of Microgrid Electricity Demand Using a Hybrid Artificial Intelligence Model. Processes” 7 (6), 320. doi:10.3390/pr7060320

[7] Ngarambe et al, 2020; Ngarambe, J., Yun, G. Y., and Santamouris, M. (2020). “The Use of Artificial Intelligence (AI) Methods in the Prediction of thermal comfort in Buildings: Energy Implications of AI-Based thermal comfort Controls. Energy and Buildings” 211, 109807. doi:10.1016/j.enbuild.2020.109807

[8] W. Strielkowski, M. Dvořák, P. Rovný, E. Tarkhanova, “5G Wireless Networks in the Future Renewable Energy Systems. In Frontiers in Energy research” , vol. 9, 2021, DOI: 10.3389/fenrg.2021.714803

Contact details:

Rola NAJA – rnaja@ece.fr

Aakash SONI – aakash.soni@ece.fr