-
[I-JEPA] Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture*STA9132/Essays 2025. 11. 23. 15:54
https://ai.meta.com/blog/yann-lecun-ai-model-i-jepa/
The first AI model based on Yann LeCun’s vision for more human-like AI
Illustrating how the predictor learns to model the semantics of the world. For each image, the portion outside of the blue box is encoded and given to the predictor as context. The predictor outputs a representation for what it expects to be in the region
ai.meta.com
https://ai.meta.com/blog/yann-lecun-advances-in-ai-research/
Yann LeCun on a vision to make AI systems learn and reason like animals and humans
The cost module computes a single scalar output that predicts the level of discomfort of the agent. It is composed of two submodules: the intrinsic cost, which is hard-wired and immutable (not trainable), and computes the immediate discomfort (such as dama
ai.meta.com
'*STA9132 > Essays' 카테고리의 다른 글
Essay 6 - REPA (0) 2025.11.24 [REPA] Representation Alignment for Generation: Training DiT (0) 2025.11.23 [MAE] Masked Autoencoders Are Scalable Vision Learners (0) 2025.11.23 [MoCo v3] An Empirical Study of Training Self-Supervised Vision Transformers (0) 2025.11.23 [SimCLR] A Simple Framework for Contrastive Learning of Visual Representations (0) 2025.11.23