diff --git a/_data/citations.yaml b/_data/citations.yaml index cd7e028..987eb5c 100644 --- a/_data/citations.yaml +++ b/_data/citations.yaml @@ -4,6 +4,9 @@ title: Evaluating Neighbor Explainability for Graph Neural Networks authors: - Oscar Llorente Gonzalez + - Rana Fawzy + - Jared keown + - Michal Horemuz - "P\xE9ter Vaderna" - "S\xE1ndor Laki" - "Roland Kotrocz\xF3" @@ -11,7 +14,7 @@ - "J\xE1nos M\xE1rk Szalai-Gindl" publisher: arXiv date: '2023-11-14' - link: https://doi.org/gs9w7h + link: https://doi.org/10.48550/arxiv.2311.08118 type: paper image: images/light.jpg description: Explainability in Graph Neural Networks (GNNs) is a new field growing @@ -36,13 +39,15 @@ link: EricssonResearch/gnn-neighbors-xai plugin: sources.py file: sources.yaml + + - id: doi:10.48550/arXiv.2310.19573 title: Model Uncertainty based Active Learning on Tabular Data using Boosted Trees authors: - Sharath M Shankaranarayana publisher: arXiv date: '2023-10-30' - link: https://doi.org/gs9w7j + link: https://doi.org/10.48550/arxiv.2310.19573 type: paper image: images/night.jpg description: Supervised machine learning relies on the availability of good labelled @@ -60,3 +65,38 @@ link: https://arxiv.org/abs/2310.19573 plugin: sources.py file: sources.yaml + + +- id: doi.org/10.48550/arXiv.2309.12913 + title: "A matter of attitude: Focusing on positive and active gradients to boost saliency maps" + date: '2023-09-22' + type: paper + image: images/space.jpg + authors: + - Oscar Llorente Gonzalez + - Jaime Boal + - "Eugenio F. S\xE1nchez-\xDAbeda" + publisher: arXiv + description: ESaliency maps have become one of the most widely used interpretability + techniques for convolutional neural networks (CNN) due to their simplicity and + the quality of the insights they provide. However, there are still some doubts + about whether these insights are a trustworthy representation of what CNNs use + to come up with their predictions. This paper explores how rescuing the sign of + the gradients from the saliency map can lead to a deeper understanding of multi-class + classification problems. Using both pretrained and trained from scratch CNNs we + unveil that considering the sign and the effect not only of the correct class, + but also the influence of the other classes, allows to better identify the pixels + of the image that the network is really focusing on. Furthermore, how occluding + or altering those pixels is expected to affect the outcome also becomes clearer. + tags: + - GAI Lab + - Comillas Pontifical University (ICAI) + buttons: + - type: paper + text: Manuscript + link: https://arxiv.org/abs/2309.12913 + - type: github + text: Source Code + link: osllogon/positive_active_saliency_maps + plugin: sources.py + file: sources.yaml diff --git a/_data/projects.yaml b/_data/projects.yaml index b4902a0..8e88a78 100644 --- a/_data/projects.yaml +++ b/_data/projects.yaml @@ -1,8 +1,8 @@ - title: GAI Lab subtitle: Geometric Artficial Intelligence group: featured - image: images/sa.png - link: https://github.com/ + image: images/gai_lab.png + link: https://ericsson.github.io/cognitive-labs/publications/ description: Research Lab focused on Geometric AI, specifically in GNNs, Explainability in Geometric ML, Graph Generation and Temporal GNNs. tags: - GAI Lab @@ -10,8 +10,8 @@ - title: MLR Lab subtitle: Machine Learning and Reasoning group: featured - image: images/mlr_icon.png - link: https://github.com/ + image: images/mlr_lab.png + link: https://ericsson.github.io/cognitive-labs/publications/ description: Research Lab focused on Optimization, Reinforcement Learning, Distributed Learning, Catastrophic Fogetting and Active Learning. tags: - MLR Lab @@ -19,26 +19,8 @@ - title: FAI Lab subtitle: Foundational Artficial Intelligence group: featured - image: images/fai_icon.png - link: https://github.com/ + image: images/fai_lab.png + link: https://ericsson.github.io/cognitive-labs/publications/ description: Research Lab focused on Generative Artificial Intelligence, Catatrophic Forgetting and Long-term Forecasting with Temporal Model. tags: - - FAI Lab - -- title: Cool Web App - subtitle: a subtitle - image: images/photo.jpg - link: https://github.com/ - description: Lorem ipsum _dolor sit amet_, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. - repo: greenelab/lab-website-template - tags: - - software - -- title: Cool Web Server - subtitle: a subtitle - image: images/photo.jpg - link: https://github.com/ - description: Lorem ipsum _dolor sit amet_, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. - repo: greenelab/lab-website-template - tags: - - software + - FAI Lab \ No newline at end of file diff --git a/_data/sources.yaml b/_data/sources.yaml index 3069508..e58f041 100644 --- a/_data/sources.yaml +++ b/_data/sources.yaml @@ -4,6 +4,9 @@ image: images/light.jpg authors: - Oscar Llorente Gonzalez + - Rana Fawzy + - Jared keown + - Michal Horemuz - "P\xE9ter Vaderna" - "S\xE1ndor Laki" - "Roland Kotrocz\xF3" @@ -33,4 +36,25 @@ - type: paper text: Manuscript link: https://arxiv.org/abs/2310.19573 - + + +- id: doi.org/10.48550/arXiv.2309.12913 + + date: '2023-09-22' + type: paper + image: images/space.jpg + authors: + - Oscar Llorente Gonzalez + - Jaime Boal + - Eugenio F. Sánchez-Úbeda + description: ESaliency maps have become one of the most widely used interpretability techniques for convolutional neural networks (CNN) due to their simplicity and the quality of the insights they provide. However, there are still some doubts about whether these insights are a trustworthy representation of what CNNs use to come up with their predictions. This paper explores how rescuing the sign of the gradients from the saliency map can lead to a deeper understanding of multi-class classification problems. Using both pretrained and trained from scratch CNNs we unveil that considering the sign and the effect not only of the correct class, but also the influence of the other classes, allows to better identify the pixels of the image that the network is really focusing on. Furthermore, how occluding or altering those pixels is expected to affect the outcome also becomes clearer. + tags: + - GAI Lab + - Comillas Pontifical University (ICAI) + buttons: + - type: paper + text: Manuscript + link: https://arxiv.org/abs/2309.12913 + - type: github + text: Source Code + link: osllogon/positive_active_saliency_maps diff --git a/_members/abdelrahman.md b/_members/abdelrahman.md index 799c063..c3db158 100644 --- a/_members/abdelrahman.md +++ b/_members/abdelrahman.md @@ -9,6 +9,6 @@ links: email: abdelrahman.el.antably@ericsson.com --- -# Abdelrahman +# Abdelrahman El Antably Abdelrahman is a Data Scientist at Ericsson with more than four years of experience in data analysis/data management field. He is currently part of the MMEA AI and Analytics team, as well as the Cognitive Software R&D working with the Uplink Optimizer team. Concurrently, Abdelrahman is pursuing a Master's degree at Cairo University, specializing in AI applications in Telecom, after completing the pre-master's studies. The academic journey began at Ain Shams University where he obtained his Bachelor's degree in Computer Science and Bioinformatics. Abdelrahman worked previously at Udacity, where he had the privilege of educating students on data analysis courses and assisting them in applying data analysis concepts to solve real-world problems. diff --git a/_members/alex-geogloman.md b/_members/alex-geogloman.md index ffaffb3..1b47065 100644 --- a/_members/alex-geogloman.md +++ b/_members/alex-geogloman.md @@ -1,6 +1,6 @@ --- name: Alex Geogloman -image: images/members/leaders/alex.jpeg +image: images/members/leaders/alex.png role: phd description: Data Architecture Manager lab: leaders @@ -9,4 +9,6 @@ links: email: alexandru.geogloman@ericsson.com --- -The best one when it comes to the connection between Radio and AI. +# Alexandru Geogloman + +Alexandru Geogloman is a seasoned professional with over a decade of rich experience in industrial automation and software engineering. His background includes significant leadership roles in sales, technical sales, engineering learning and development, and software development. His expertise lies in leadership, strategy, problem-solving, business operations, and fostering continuous improvement. With an educational foundation that includes an MBA from the University of Hull and degrees in Electrical Engineering from Politehnica University of Bucharest, Alexandru champions the principle of continuous learning. diff --git a/_members/arindam-dey.md b/_members/arindam-dey.md index 32fa111..bfc2e05 100644 --- a/_members/arindam-dey.md +++ b/_members/arindam-dey.md @@ -10,4 +10,6 @@ links: email: arindam.d.dey@ericsson.com --- -Arindam has been associated with the cellular wireless industry for 24 years. As an Network Optimization Expert, he has worked for operators in India and abroad alike. He carries a rich experience in 2G, 3G, 4G Optimization and planning across multi-vendor networks. He joined Ericsson in 2013 as a first line leader, and since then has managed optimization teams in multiple markets. He has been also involved with L&CD , designing curriculum for enhancing Programing, Public Cloud and Machine Learning skills. He joins us from the erstwhile BMAS SA NDO SL-OPT, where he was leading multiple automation initiatives and solutions, aligned with the organization strategy. +# Arimdam Dey + +Arindam has been associated with the cellular wireless industry for 24 years. As a Network Optimization Expert, he has worked for operators in India and abroad alike. He has a rich experience in 2G, 3G, 4G Optimization and planning across multivendor networks. He joined Ericsson in 2013 as a first-line leader, and since then has managed optimization teams in multiple markets. Furthermore, he has been also involved with L&CD, designing a curriculum for enhancing Programming, Public Cloud and Machine Learning skills. He joins us from the erstwhile BMAS SA NDO SL-OPT, where he was leading multiple automation initiatives and solutions, aligned with the organization's strategy. diff --git a/_members/juan-ramiro.md b/_members/juan-ramiro.md index a7014ec..acf90c2 100644 --- a/_members/juan-ramiro.md +++ b/_members/juan-ramiro.md @@ -5,9 +5,12 @@ role: phd description: AI Research Manager lab: leaders links: - email: raul.martin@ericsson.com + linkedin: juan-ramiro + email: juan.ramiro@ericsson.com --- +# Juan Ramiro + Juan Ramiro is the AI Research Manager within Ericsson Cognitive Software Engineering, where he leads R&D activities with mid and long-term horizon, involving exploration of new technologies, forward-looking concepts and disruptive network optimization methodologies. With over 15 years of experience in the telecom industry, he has held positions related to R&D, strategy, marketing and technical sales. Juan holds a Telecom Engineering degree from Malaga University, a PhD in Electrical and Electronic Engineering from Aalborg University, an Executive MBA from San Telmo Business School and an Executive Degree in Big Data & Business Analytics from EOI. He is also an Honorary Associate Professor at Malaga University, as well as the co-author of a book on Self-Organizing Networks diff --git a/_members/judit-badics.md b/_members/judit-badics.md index 433e797..7b32d12 100644 --- a/_members/judit-badics.md +++ b/_members/judit-badics.md @@ -9,4 +9,6 @@ links: email: judit.badics@ericsson.com --- +# Judit Badics + A data scientist with a strong mathematical and game theoretic background who uses a multidisciplinary way of thinking to understand how ML/DL algorithms work. diff --git a/_members/katalin.md b/_members/katalin.md index 0518dae..3a8db44 100644 --- a/_members/katalin.md +++ b/_members/katalin.md @@ -5,10 +5,13 @@ role: scientist lab: gai-lab links: linkedin: katalin-hajdu-szucs + github: HSzKatalin google-scholar: https://scholar.google.com/citations?user=mbbzmg8AAAAJ&hl=en email: katalin.hajdu-szucs@ericsson.com --- +# Katalin Hajdú-Szücs + Katalin gained her bachelor’s degree in mathematics in Budapest and her master’s degree in computer science as an EIT ICT Labs student in Trento. She continued her studies in the Doctoral School of Eötvös Loránd University at the Department of Information Systems. Currently, she is writing her dissertation in the topic of graph neural network interpretability. Her other experiences involve optical character recognition, loan default prediction, detection and classification of sound events and radio network data analytics. -She joined Ericsson in 2023 after a two-year long collaboration in the topic of graph neural network explainability. Today she is working on performance diagnostics, but she will be joining performance optimizers in the near future. +She joined Ericsson in 2023 after a two-year-long collaboration on the topic of graph neural network explainability. Today she is working on performance diagnostics, but she will be joining performance optimizers in the near future. diff --git a/_members/luis-diaz.md b/_members/luis-diaz.md index eea3a68..b3b1f5b 100644 --- a/_members/luis-diaz.md +++ b/_members/luis-diaz.md @@ -3,13 +3,15 @@ name: Luis Diaz image: images/members/gai-lab/luis.png role: scientist description: Head of FAI Lab -lab: gai-lab, fai-lab, board +lab: fai-lab, board links: linkedin: luis-felipe-diaz-9831a239 github: luisfelipe2105 email: luis.diaz@ericsson.com --- +# Luis Diaz + With over 10 years of experience in the telecommunications industry, I am a passionate data scientist who thrives on solving complex problems and creating innovative solutions. Currently, I work as a Data Scientist R&D at Ericsson, where I actively research and develop advanced machine learning models for the optimization of Radio Access Networks (RAN). My core competencies include Generative AI, Graph Neural Networks (GNN), Convolutional Neural Networks (CNN), and Neural Networks for tabular data classification. These techniques enable me to leverage diverse data sources and craft dynamic solutions that enhance and optimize RAN performance. As a result, I contribute significantly to the development of our state-of-the-art RAN optimization products. Additionally, I am expanding my expertise in applying AI to Cybersecurity, a field that offers exciting challenges and opportunities. diff --git a/_members/mariam-elmasry.md b/_members/mariam-elmasry.md index c42859b..20ac888 100644 --- a/_members/mariam-elmasry.md +++ b/_members/mariam-elmasry.md @@ -4,7 +4,11 @@ image: images/members/gai-lab/mariam.jpeg role: scientist lab: gai-lab links: + linkedin: mariam-elmasry- + github: mhelmasr email: mariam.elmasry@ericsson.com --- -First-league Research Scientist in applied Reinforcement Learning to cellular networks. +# Mariam Elmasry + +Mariam is a highly motivated passionate Experienced Data Scientist with a Master’s Degree in Advanced Analytics. She has eight years of industry experience with extensive AI, statistical and analytics background. She possesses the knowledge, experience, and techniques necessary to deliver sophisticated AI solutions. A master programmer in Python who led several team projects administrated on Azure DevOps in a variety of fields like Financial, R&D, FMCG, SAAS, etc. diff --git a/_members/mayar-hefny.md b/_members/mayar-hefny.md index 655459e..971fd72 100644 --- a/_members/mayar-hefny.md +++ b/_members/mayar-hefny.md @@ -10,5 +10,7 @@ links: email: mayar.hefnyn@ericsson.com --- +# Mayar Hefny + Mayar is an Experienced Data Scientist with more than 5 years of experience in developing AI-powered software at Ericsson. She's also pursuing a Master's degree in Machine Learning at Georgia Tech. She has contributed to multiple software products globally and has experience in different areas of the ML ecosystem. She has also worked on Trustworthy AI research and development at Ericsson and contributed to 2 patents. Her main research interests are Reinforcement Learning, Explainable RL and Quantum Machine Learning. diff --git a/_members/oscar-llorente-gonzalez.md b/_members/oscar-llorente-gonzalez.md index 1b49547..bb023e8 100644 --- a/_members/oscar-llorente-gonzalez.md +++ b/_members/oscar-llorente-gonzalez.md @@ -4,8 +4,6 @@ image: images/members/gai-lab/oscar.jpeg role: scientist description: Head of GAI Lab lab: gai-lab, board -aliases: - - Oscar Llorente Gonzalez links: linkedin: osllogon github: osllogon diff --git a/_members/philipp-frank.md b/_members/philipp-frank.md index 181efa7..ac069b4 100644 --- a/_members/philipp-frank.md +++ b/_members/philipp-frank.md @@ -5,7 +5,8 @@ role: phd description: Head of Research lab: leaders links: - email: raul.martin@ericsson.com + linkedin: philipp-frank-b557263b + email: philipp.frank@ericsson.com --- -The best one when it comes to the connection between Radio and AI. +# Philippp Frank diff --git a/_members/rana-fawzy.md b/_members/rana-fawzy.md index a598207..7d6eab9 100644 --- a/_members/rana-fawzy.md +++ b/_members/rana-fawzy.md @@ -11,6 +11,4 @@ links: # Rana Fawzy -Senior Data Scientist with over five years of expertise in leveraging machine learning, deep learning and statistical -modeling to address complex business challenges. Proficient in transforming data into strategic opportunities for -actionable insights. +Data scientist with over five years of experience, focusing on developing and applying machine learning and deep learning algorithms, along with statistical modeling and data analysis, specifically in the telecom industry. I’m always on the lookout for the newest and most cutting-edge algorithms to keep our work at the forefront of technology. diff --git a/_members/raul-martin.md b/_members/raul-martin.md index a14bd33..6316eef 100644 --- a/_members/raul-martin.md +++ b/_members/raul-martin.md @@ -5,7 +5,10 @@ role: phd description: AI Lead lab: leaders links: + linkedin: raul-martín-cuerdo-b6a53913 email: raul.martin@ericsson.com --- -The best one when it comes to the connection between Radio and AI. +# Raul Martin + +Raúl Martín Cuerdo is ML and AI product development leader in Ericsson Cognitive Software Engineering, helping to realize ideas and concepts into new AI powered software applications during the last 6 years. With more than 20 years of experience in telecom industry, he has held positions as customer technical project lead for relevant CSPs in different companies. He holds a master’s degree in telecommunications engineering from the Universidad Politécnica de Madrid. diff --git a/_members/sarah-johnson.md b/_members/sarah-johnson.md deleted file mode 100644 index b0d97b4..0000000 --- a/_members/sarah-johnson.md +++ /dev/null @@ -1,11 +0,0 @@ ---- -name: Sarah Johnson -image: images/photo.jpg -description: Lead Programmer -role: programmer -links: - email: sarah.johnson@gmail.com - twitter: sarahjohnson ---- - -Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. diff --git a/_members/sharath-x.md b/_members/sharath-x.md index 6da0321..0391df1 100644 --- a/_members/sharath-x.md +++ b/_members/sharath-x.md @@ -3,9 +3,13 @@ name: Sharath X image: images/members/mlr-lab/sharath.png role: scientist description: Head of MLR Lab -lab: mlr-lab, gai-lab, fai-lab, board +lab: mlr-lab, fai-lab, board links: + linkedin: sharath-m-s + github: sharath-sms + google-scholar: https://scholar.google.com/citations?hl=en&user=HMQHVhoAAAAJ&view_op=list_works&sortby=pubdate email: sharath.x@ericsson.com + orcid: 0000-0003-3482-4887 --- # Sharath X diff --git a/_members/veera.md b/_members/veera.md index c990906..1d47584 100644 --- a/_members/veera.md +++ b/_members/veera.md @@ -4,7 +4,8 @@ image: images/members/gai-lab/veera.jpeg role: scientist lab: gai-lab links: + linkedin: veerarc email: chikka.veera.raghavendra@ericsson.com --- -The best one when it comes to the connection between Radio and AI. +# Chikka Veera \ No newline at end of file diff --git a/_members/ziad-osman.md b/_members/ziad-osman.md index 81ac9d4..46d8085 100644 --- a/_members/ziad-osman.md +++ b/_members/ziad-osman.md @@ -9,7 +9,9 @@ links: email: ziad.osman@ericsson.com --- -Ziad is the proprietor of an exquisite record of achieving industrial & research goals & products for worldwide top-ranked universities’ product development and research labs under the umbrella of top tier leading technology enterprises. +# Ziad Osman + +Ziad is the proprietor of an exquisite record of achieving industrial & research goals & products for worldwide top-ranked universities’ product development and research labs under the umbrella of top-tier leading technology enterprises. He is a fervent researcher in quantum computing and its symbiotic association with machine learning and data science. Moreover, he has a Master of science degree (M.Sc.) in electronics and communications engineering in “Enhancing quantum computing using machine learning techniques”; one of the first theses –at the time- to make use of that paradigm in the rapport between quantum computing and machine learning. -He obtained bachelor’s degree (B.Sc.) in electronics and communications engineering from AASTMT with ABET certified coursework and top rankings among his class. Former Teaching assistant/researcher at AASTMT college of engineering (electronics and communications engineering and computer engineering) and college of computer science. +He obtained a bachelor’s degree (B.Sc.) in electronics and communications engineering from AASTMT with ABET-certified coursework and top rankings among his class. Former Teaching assistant/researcher at AASTMT College of Engineering (electronics and communications engineering and computer engineering) and college of computer science. diff --git a/blog/index.md b/blog/index.md index b86f359..85492bf 100644 --- a/blog/index.md +++ b/blog/index.md @@ -1,11 +1,11 @@ --- title: Blog -nav: - order: 4 - tooltip: Musings and miscellany +# nav: +# order: 4 +# tooltip: Musings and miscellany --- -# {% include icon.html icon="fa-solid fa-feather-pointed" %}Blog +# {% include icon.html icon="fa-solid fa-feather-pointed" %}Blog2 Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. diff --git a/images/fai_icon.png b/images/fai_lab.png similarity index 100% rename from images/fai_icon.png rename to images/fai_lab.png diff --git a/images/sa.png b/images/gai_lab.png similarity index 100% rename from images/sa.png rename to images/gai_lab.png diff --git a/images/members/leaders/alex.jpeg b/images/members/leaders/alex.jpeg deleted file mode 100644 index a5a22d9..0000000 Binary files a/images/members/leaders/alex.jpeg and /dev/null differ diff --git a/images/members/leaders/alex.png b/images/members/leaders/alex.png new file mode 100644 index 0000000..dfa089d Binary files /dev/null and b/images/members/leaders/alex.png differ diff --git a/images/mlr_icon.png b/images/mlr_lab.png similarity index 100% rename from images/mlr_icon.png rename to images/mlr_lab.png diff --git a/projects/index.md b/labs/index.md similarity index 92% rename from projects/index.md rename to labs/index.md index a878015..8bb049e 100644 --- a/projects/index.md +++ b/labs/index.md @@ -2,7 +2,7 @@ title: Labs nav: order: 2 - tooltip: Software, datasets, and more + tooltip: Our research organization --- # {% include icon.html icon="fa-solid fa-graduation-cap" %} Labs diff --git a/research/index.md b/publications/index.md similarity index 64% rename from research/index.md rename to publications/index.md index f858b7d..e804cca 100644 --- a/research/index.md +++ b/publications/index.md @@ -7,7 +7,7 @@ nav: # {% include icon.html icon="fa-solid fa-scroll" %} Publications -On this page, all the publications from Ericsson Cognitive Network Solutions are listed, with links to the manuscript and the code to ensure reproducibility. We have also provided tags, so you can check our collaborators for each publication. +On this page, all the publications from Ericsson Cognitive Labs are listed, with links to the manuscript and the code to ensure reproducibility. We have also provided tags, so you can check our collaborators for each publication. ## Highlighted