A **Manifold** is a lower-[[Dimensions|dimensional]] surface that has been embedded in a higher dimensional space. A torus exists in 3-dimensional space. It's confusing to try to think of the points on the surface of a donut... but you can address them uniquely using the two angles. These two angles are your manifold for your 3-dimensional object. Same goes for a sphere. If you discover a manifold that the data instances you care about inhabit, you can do [[Machine Learning]] training on just that manifold rather than the entire high-dimensional space. **** # More ## Source - Grad School