Contrastive Learning News Today

Researchers developed Prov-GigaPath, a whole-slide pathology foundation model using a novel vision transformer architecture. The model demonstrates superior performance in mutation prediction, cancer subtyping, and vision-language tasks. It leverages large-scale real-world data from over 30,000 patients to enhance clinical diagnostics.

Stick a camera on a child, then feed what it captures to an AI, and it almost works.

They trained an AI through the eyes of a baby

Imagine if an artificial intelligence (AI) model could learn language just like a child does—by seeing and hearing the world through their eyes and ears.

Researchers at NYU trained a CVCL model to link words with visual cues. The model was trained on data from a headset worn by a toddler and containing about 60 hours of footage.

Research establishes a computational basis to study how children begin to speak, connecting what they see with the auditory stimuli they receive from adults

Graph embedding aims to embed the information of graph data into low-dimensional representation space. Prior methods generally suffer from an imbalance of preserving structural information and node features due to their pre-defined inductive biases, leading to unsatisfactory generalization performance. In order to preserve the maximal information, graph contrastive learning (GCL) has become a prominent technique for learning discriminative embeddings. However, in contrast with graph-level embedd...

MIT researchers have developed StableRep, an AI training method using synthetic images generated by text-to-image models, which surpasses traditional training on real images. The approach leverages multi-positive contrastive learning, promising more efficient, less biased, and resource-conscious machine learning development.

In intelligent systems, knowledge tracing (KT) plays a vital role in providing personalized education. Existing KT methods often rely on students' learning interactions to trace their knowledge states by predicting future performance on the given questions. While deep learning-based KT models have achieved improved predictive performance compared with traditional KT models, they often lack interpretability into the captured knowledge states. Furthermore, previous works generally neglect the...

A professionally curated list of papers, tutorials, books, videos, articles and open-source libraries etc for Out-of-distribution detection, robustness, and generalization - GitHub - continuousml/Awesome-Out-Of-Distribution-Detection: A professionally curated list of papers, tutorials, books, videos, articles and open-source libraries etc for Out-of-distribution detection, robustness, and generalization