How deeply does Math and Architectures of Deep Learning cover hypersurface?

Explains geometric intuition, linear and nonlinear forms, and sign-based side assignment, but stops short of training algorithms.

Hypersurface

As explained in Math and Architectures of Deep Learning

A surface of arbitrary dimension.

A hypersurface is a surface of arbitrary dimension. In classification, it forms a decision boundary: data points on different sides belong to different classes. In high-dimensional input spaces, the boundary separates clusters of vectors representing classes. section 3.1.2

The boundary's geometry determines how classes are separated:

A hyperplane provides a linear separator when class clusters can be divided by a flat boundary. section 3.1.2 section 7.2.2
A curved hypersurface, such as a hypersphere, handles more complicated class arrangements that a hyperplane cannot separate. section 3.1.2 section 7.2.2
For a boundary described by φ(x) = 0, points with φ(x) = 0 lie on the surface, while positive and negative values identify opposite sides. section 7.2.2

Car–giraffe and horse–zebra classification

Rasterized images become points in a high-dimensional space. Car and giraffe points can be separated with a hyperplane, whereas horse and zebra points may require a curved hypersphere. Training determines the boundary parameters, such as a hyperplane's orientation and position or a hypersphere's center and radius. section 3.1.2

How Math and Architectures of Deep Learning covers it

Cover of Math and Architectures of Deep Learning

liveBook manning.com

Read Math and Architectures of Deep Learning on liveBook Get the book at manning.com

Related concepts

autoencoderused by · An autoencoder is a neural network that encodes input data into an internal representation and decodes it to reconstruct the input or an approximation.
decision boundaryequivalent to · A decision boundary is the locus in input or feature space that separates regions assigned different classes or decision outcomes by a predictive model.
hyperplanecontrasts with · A hyperplane is a flat affine subspace with dimension one less than its ambient space, capable of dividing that space into regions.
loss surfaceequivalent to · A loss surface is a function that assigns a loss value to each possible setting of a model’s parameters.
surfaceincludes · A surface is a two-dimensional geometric set, often embedded in three-dimensional space or representing a boundary or relation.

This explainer was generated by Manning's knowledge graph from the text of Math and Architectures of Deep Learning. The section links point to the parts of the book it draws on.