![](Screen%20Shot%202022-12-10%20at%208.23.22%20AM.png) > Deep learning is about *relating samples to each other* through a **learned distance function**. ![](Screen%20Shot%202022-12-10%20at%208.38.07%20AM.png) ![](Screen%20Shot%202022-12-10%20at%208.39.41%20AM.png) ![](Screen%20Shot%202022-12-10%20at%208.40.33%20AM.png) See the [Manifold Hypothesis](Manifold%20Hypothesis.md). ![](Screen%20Shot%202022-12-11%20at%208.37.00%20AM.png) ![](Screen%20Shot%202022-12-11%20at%208.37.29%20AM.png) ![](Screen%20Shot%202022-12-11%20at%208.37.43%20AM.png) ![](Screen%20Shot%202022-12-11%20at%208.38.01%20AM.png) ![](Screen%20Shot%202022-12-11%20at%208.38.13%20AM.png) ![](Screen%20Shot%202022-12-11%20at%208.38.30%20AM.png) --- Date: 20221210 Links to: Tags: #review References: * [Francois Chollet, Melanie Mitchell, Christian Szegedy · Abstraction & Reasoning in AI systems: Modern Perspectives · SlidesLive](https://slideslive.com/38935790/abstraction-reasoning-in-ai-systems-modern-perspectives) * [Francois Chollet | Why abstraction is the key to intelligence, and what we’re still missing - YouTube](https://www.youtube.com/watch?v=3Nxe7J07TQY)