Track synthesis, materials and catalysis as it is published.
Search papers, labs, and topics across Lattice.
This paper redefines topological expressivity for neural networks beyond binary classification by analyzing the Betti numbers of the network's output space, $\beta(\Phi(X))$. It establishes constructive upper bounds for the expressivity of ReLU networks, demonstrating that specific architectures can represent spaces with arbitrary Betti numbers, including wedge sums of spheres. The work culminates in a conjecture for a tight bound for networks representing spheres, offering a more precise understanding of neural network function approximation capabilities.
Neural networks can precisely encode arbitrary topological spaces, with their expressivity directly tied to the Betti numbers of their output, not just decision boundaries.
The expressivity of a neural network $\Phi$ describes the class of functions that a network can express. The term topological expressivity refers to a family of measures particularly suited to understand binary classification and is often formulated in terms of the Betti numbers of a decision boundary $\beta (\Phi^{-1}(0))$. Despite its popularity, this measure is not as useful for regression and other problems that do not involve a decision boundary. Instead, we propose to study topological expressivity in terms of the Betti numbers of the output of a neural network $\beta(\Phi(X))$. We formally prove that a ReLU network with architecture $(d,\sum\limits_{k=1}^{d } \beta_k(3k+2),\sum\limits_{k=1}^{d}\beta_k(k+1))$ can express spaces homotopy equivalent to wedge sums of spheres with arbitrary Betti numbers. This gives upper bounds for any arbitrary connected ($\beta_0=1$) sequence of Betti numbers, but by constructing spaces homotopy equivalent to products of $N$ spheres of dimension $d_1,...,d_N$ we achieve much better bounds $(\sum\limits_{n=1} d_n, 3\sum\limits_{n=1}\limits^N d_n + 2N, \sum\limits_{n=1}\limits^N d_n + N) $ for Betti numbers given by $\prod_n (1+x^{d_n})$. Finally we conjecture a much stronger tight bound which gives the architectural bound $(d,d+1,d+1)$ for a space homeomorphic to the sphere.