A5038662752- Recommender Systems and Techniques
- Topic Modeling
- Machine Learning in Healthcare
- Advanced Bandit Algorithms Research
- Sentiment Analysis and Opinion Mining
- Advanced Graph Neural Networks
- Machine Learning and Data Classification
- Traffic Prediction and Management Techniques
- Diabetes Management and Research
- Data Stream Mining Techniques
- Image Retrieval and Classification Techniques
- Privacy-Preserving Technologies in Data
- Stochastic Gradient Optimization Techniques
- Data Mining Algorithms and Applications
- Machine Learning in Materials Science
- Multimodal Machine Learning Applications
- Retinal Imaging and Analysis
- Genetics, Bioinformatics, and Biomedical Research
- Computational Drug Discovery Methods
- Natural Language Processing Techniques
- Generative Adversarial Networks and Image Synthesis
Telecom SudParis
2019-2022
Institut Polytechnique de Paris
2019-2022
Orange (France)
2022
Nvidia (United Kingdom)
2021
Centre National de la Recherche Scientifique
2019
Télécom Paris
2019
Institut Mines-Télécom
2019
Much of the recent progress in sequential and session-based recommendation has been driven by improvements model architecture pretraining techniques originating field Natural Language Processing. Transformer architectures particular have facilitated building higher-capacity models provided data augmentation training which demonstrably improve effectiveness recommendation. But with a thousandfold more research going on NLP, application transformers for understandably lags behind. To remedy...
Recently, large language models (LLMs) have exhibited significant progress in understanding and generation. By leveraging textual features, customized LLMs are also applied for recommendation demonstrate improvements across diverse scenarios. Yet the majority of existing methods perform training-free that heavily relies on pretrained knowledge (e.g., movie recommendation). In addition, inference is slow due to autoregressive generation, rendering less effective real-time recommendation. As...
Objective The objective of this article was to compare the performances health care-associated infection (HAI) detection between deep learning and conventional machine (ML) methods in French medical reports. Methods corpus consisted different types reports (discharge summaries, surgery reports, consultation etc.). A total 1,531 text documents were extracted deidentified three university hospitals. Each them labeled as presence (1) or absence (0) HAI. We started by normalizing records using a...
Session-based recommendation is an important task for e-commerce services, where a large number of users browse anonymously or may have very distinct interests different sessions. In this paper we present one the winning solutions Recommendation SIGIR 2021 Workshop on E-commerce Data Challenge. Our solution was inspired by NLP techniques and consists ensemble two Transformer architectures - Transformer-XL XLNet trained with autoregressive autoencoding approaches. To leverage most rich...
Newcomers to recommender systems often face challenges related their lack of understanding how these operate in real life. In most online content this topic, the focus is on models and algorithms that score items based user's preferences. However, model alone does not comprise everything needed for serving optimized meet company's business objectives. An industry-standard system involves a number steps, including data preprocessing, defining training models, as well filtering logic serving....
In this paper we present the novel aspects of our 15th place solution to RecSys Challenge 2019 which are focused on acceleration feature generation and model training time. final sped up by a factor 15.6x, from workflow 891.8s (14m52s) 57.2s, through combination RAPIDS.AI cuDF library for preprocessing, custom batch dataloader, LAMB extreme sizes, an update kernel responsible calculating embedding gradient in PyTorch. Using also accelerated 9.7x performing computations GPU, reducing time...
Synthetic data and simulators have the potential to markedly improve performance robustness of recommendation systems. These approaches already had a beneficial impact in other machine-learning driven fields. We identify discuss key trade-off between fidelity privacy past work on synthetic for For important use case predicting algorithm rankings real from data, we provide motivation current successes versus limitations. Finally outline number exciting future directions systems that believe...
In recent years, several deep learning-based algorithms have been proposed for recommendation systems while its adoption in industry deployments steeply growing. particular, NLP-inspired approaches successfully adapted sequential and session-based problems, which are important many domains like e-commerce, news streaming media. this regard, hands-on tutorial will offer to the participants: (I) an introduction on main concepts recommendation, (II) how build, train evaluate a model based RNN...
Artificial Intelligence models encoding biology and chemistry are opening new routes to high-throughput high-quality in-silico drug development. However, their training increasingly relies on computational scale, with recent protein language (pLM) hundreds of graphical processing units (GPUs). We introduce the BioNeMo Framework facilitate AI across GPUs. Its modular design allows integration individual components, such as data loaders, into existing workflows is open community contributions....
Session-based recommendation is an important task for domains like e-commerce, that suffer from the user cold-start problem due to anonymous browsing and which users preferences might change considerably over time. The RecSys Challenge 2022, organized by Dressipi, focused on session-based fashion e-commerce domain. In this paper, NVIDIA RAPIDS Merlin teams present their solution placed 3rd in challenge. Among most effective techniques we found sessions augmentation ensembling a very diverse...
Industrial recommender systems are made up of complex pipelines requiring multiple steps including feature engineering and preprocessing, a retrieval model for candidate generation, filtering, store query, ranking scoring, an ordering stage. These need to be carefully deployed as set, coordination during their development deployment. Data scientists, ML engineers, researchers might focus on different stages systems, however they share common desire reduce the time effort searching combining...
Recommender Systems (RecSys) are the engine of modern internet and catalyst for human decisions. The goal a recommender system is to generate relevant recommendations users from collection items or services that might interest them. Building recommendation challenging because it requires multiple stages (item retrieval, filtering, ranking, ordering) work together seamlessly efficiently during training inference. biggest challenges faced by new practitioners lack understanding around what...