Your Daily AI Research tl;dr — 2022–11–02 🧠

A new tool for in-line code suggestions, Music editing and generation, and Text-Only Training for Image Captioning!

Welcome to your official daily AI research tl;dr (often with code and news) for AI professionals where I share the most exciting papers I find daily, along with a one-liner summary to help you quickly determine if the article (and code) is worth investigating.

1️⃣ Text-Only Training for Image Captioning using Noise-Injected CLIP

They consider the task of image-captioning using only the CLIP model and additional text data at training time, and no additional captioned images, using a frozen CLIP encoder and learning a decoder to translate CLIP textual embeddings back into text.

Link to the paper: https://arxiv.org/pdf/2211.00575.pdf

2️⃣ [Sponsored] Build better machine learning models with Comet

Comet ML is a platform for tracking and visualizing machine learning models. With 2 lines of code, Comet automatically tracks your hyperparameters, environment, source code, and data. It gives you visualizations that let you compare how different models perform. It’s a powerful tool that’s already integrated with the most popular machine learning frameworks like Hugging Face, YOLOv5 and PyTorch Lightning.

Create your free Comet account today

3️⃣ SDMUSE: STOCHASTIC DIFFERENTIAL MUSIC EDITING AND GENERATION VIA HYBRID REPRESENTATION

“We propose SDMuse, a unified Stochastic Differential Music editing and generation framework, which can not only compose a whole musical piece from scratch, but also modify existing musical pieces in many ways, such as combination, continuation, inpainting, and style transferring.”

Link to the paper: https://arxiv.org/pdf/2211.00222.pdf

Join the discussion with fellow AI enthusiasts

🆕 A new tool for in-line code suggestions!

Baed on a language model, the startup Replit introduced a new programming tool, Ghostwriter, that offers in-line code suggestions and other features.

Please subscribe and share the newsletter with your AI techy friends if it was helpful and join our 4'000+ readers. Feel free to follow my weekly newsletter where I go in-depth into one of the papers shared, and our newsletter at Towards AI, sharing the most exciting news, papers, articles, and memes weekly.

Thank you for reading,

Louis