articles - 1/1 Blog article (8) articles (3) cours (6) modules (11) papers (5)

---

articles - 1/1

Quelques articles intéressants

2022-07-17

Deep Learning

• Training Deep Networks without Learning Rates Through Coin Betting

• Model Comparison and Calibration Assessment: User Guide for Consistent Scoring Functions in Machine Learning and Actuarial Practice

• On Embeddings for Numerical Features in Tabular Deep Learning <https://arxiv.org/abs/2203.05556>, TABULAR DATA MODELING VIA CONTEXTUAL EMBEDDINGS

• MLGO: a Machine Learning Guided Compiler Optimizations Framework

• Embedding Recycling for Language Models

• Language Modelling with Pixels

• XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale

• Emerging Cross-lingual Structure in Pretrained Language Models

• Word Translation Without Parallel Data

• Very Deep Convolutional Networks for Text Classification

• XTREME-S: Evaluating Cross-lingual Speech Representations

• DeepSpeed Compression: A composable library for extreme compression and zero-cost quantization

• Towards Robust Deep Learning using Entropic Losses

• Few-shot Learning with Retrieval Augmented Language Models

• Multi-instrument Music Synthesis with Spectrogram Diffusion

Machine Learning

• On the consistency of supervised learning with missing values

• Why do tree-based models still outperform deep learning on tabular data?

• Evaluating machine learning models and their diagnostic value

Ecologie

• Measuring the Carbon Intensity of AI in Cloud Instances

Un site

• paperwithcode

Librairies

• TorchArrow

Un benchmark

• MIXED PRECISION TRAINING IN PRACTICE

Autres articles

• On the Principles of Parsimony and Self-Consistency for the Emergence of Intelligence

• Shift-Curvature, SGD, and Generalization

article

Visualisation de modèles

2018-07-12

Netron : cet outil n’explique pas vraiment mais permet de visualiser des modèles de machine learning. C’est un pas dans la bonne direction, il semble assez populaire sur GitHub. Il est sans doute amené à évoluer (référencé dans Interprétabilité des modèles). VisualDL est un outil similaire dédié au deep learning.

…

article

Quelques articles à lire

2018-07-03

Lectures d’été.

…

article

---

articles - 1/1 2021-09 (4) 2022-06 (2) 2022-07 (1) 2022-09 (1) 2022-10 (1)