Knowledge sharing sheets - Predictive AI

Table of Contents (ToC)

• Knowledge sharing sheets - Predictive AI
• Table of Contents (ToC)
• General python modules
• Forecasting
• Text-to-video
• Reinforcement Learning (RL)
• Ressources

General python modules

• skrub: Prepping tables for machine learning (data wrangling)
• sklearn: machine learning in Python
• pytorch
• tensorflow
• tabPFN: Foundation Model for Tabular Data
• scikit-network
• CARTE: Repository for CARTE: Context-Aware Representation of Table Entries
• structured predictions
  • https://hi-paris.github.io/structured-predictions/_build/html/index.html
• deepdespeckling: Speckle noise seriously limits the interpretability of synthetic aperture radar (SAR) images. This package provides deep learning based despeckling methods to denoise SAR images.
• POT: Python Optimal Transport

Learning - Machine Learning in Python with scikit-learn (MOOC)

• INRIA GitHub - Scikit Learn MOOC
  • MOOC - France Université Numerique
  • Certification probal AI - AI Action Summit

Tabular Data

Github

• Github - RTDL (Research on Tabular Deep Learning)
  • Github - On Embeddings for Numerical Features in Tabular Deep Learning
• Github - Pytorch-frame
• Github - CARTE
• Github - Tabpfn time series

Articles

• Arxiv - Deep Feature Embedding for Tabular Data
• Arxiv - On Embeddings for Numerical Featuresin Tabular Deep Learning

Startup

• Neuralk AI

Forecasting

• Github - Merlion
  • Paper
• Github - chronos-forecasting (AWS)
• Github - autogluon: opensource autoML tooklit that democratizes machine learning, fast & accurate ML in 3 lines of code, supports tabular, time series, image, text, multimodal data
• Github - Nixtla: state-of-the-art time series forecasting methods that are embedded in a set of libraries and models, most famous are:
  • Github - statsforecast: Lightning fast forecasting with statistical and econometric models.
    • Documentation
  • Github - neuralforecast: Scalable and user friendly neural 🧠 forecasting algorithms.
    • Documentation
  • Github - hierarchicalforecast
    • Documentation
  • Github - TimeGPT
• Github - Darts: python module offers a consistent 𝘧𝘪𝘵() and 𝘱𝘳𝘦𝘥𝘪𝘤𝘵() interface across various forecasting models, from classical methods like ARIMA to deep learning models. This makes it easy to compare and switch between different models without needing to understand their internal workings
• Github - Prophet: designed for users with limited statistical knowledge, allowing them to easily handle seasonal effects and missing data. Its intuitive interface makes it accessible for business applications requiring quick and interpretable forecasts.
• Github - SKtime provides a scikit-learn-compatible framework for time series analysis, enabling users to apply machine learning techniques seamlessly. This integration allows for easy model tuning, validation, and deployment using familiar tools. You can also easily use embedded scikit-learn data transformers for feature generation
• Github - GluonTS is focused on probabilistic forecasting, allowing users to generate not just point forecasts but also confidence intervals.
• Github - Tabpfn time series
  • Podcast - Tabpfn
• Article - Dive into Time-Series Anomaly Detection: A Decade Review

Vision

• HF - timm
  • Github - pytorch-image-models
• Github - opencv-pythoncore: vision algorithms
• Github - torchvisioncore: vision algorithms
• Github - Pillowcore: vision algorithms
• Github - albumentationsimage: augmentations
• Github - timmpre-trained: models
• Github - scikit-imagecore: vision algorithms
• [Github -segmentation-models-pytorch] segmentation

Text-to-video

• TransPixar: pipeline designed to support various video tasks, including Text-to-RGBA Video, Image-to-RGBA Video.

Reinforcement Learning (RL)

• Keynote - Introduction to reinforcement learning

Ressources

• Github - scikit-learn (machine learning)
• Github - skrub (data wrangler)
• Github - VLMs zero to here: this series will take you on a journey from the fundamentals of NLP and Computer Vision to the cutting edge of Vision-Language Models.
• Deep Learning Book