Research map / Architectures and efficiency / RNNs and normalization
The power of deeper networks for expressing natural functions
David Rolnick, Max Tegmark
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a deep network instead of a single hidden layer | Feedforward networks approximating multivariate monomials · Multivariate monomial | neurons required: \prod_{i=1}^n (r_i + 1) -> \le \sum_{i=1}^n (7\lceil\log_2(r_i)\rceil + 4) | Improved |
| Use a deep network instead of a single hidden layer | Feedforward networks approximating sparse multivariate polynomials · Sparse multivariate polynomial | neurons required: \ge \frac{1}{c}\max_j \munif_1(q_j) -> \le \sum_j \munif(q_j) | Improved |
| Use a deep network instead of a single hidden layer | Feedforward networks approximating x^d · Univariate monomial x^d | neurons required: d + 1 -> \le 7\lceil\log_2(d)\rceil | Improved |
| Use a fixed neuron count while changing weights for increasingly accurate approximations | Feedforward networks approximating degree-d polynomials · Polynomial uniform approximation | neurons required as ε approaches zero | No clear effect |
| Use ReLU activation rather than an activation with the required nonzero Taylor coefficients | Fixed-size network approximation of nonlinear polynomials · Polynomial uniform approximation | neurons required as ε approaches zero | Hurt |
Closest papers by meaning
- Why Deep Neural Networks for Function Approximation? (ICLR 2017 · similarity 0.90)
- The Expressive Power of Neural Networks: A View from the Width (NeurIPS 2017 · similarity 0.88)
- Neural Networks with Small Weights and Depth-Separation Barriers (NeurIPS 2020 · similarity 0.88)
- A closer look at the approximation capabilities of neural networks (ICLR 2020 · similarity 0.88)
- Towards Lower Bounds on the Depth of ReLU Neural Networks (NeurIPS 2021 · similarity 0.88)
- On the Hardness of Training Deep Neural Networks Discretely (AAAI 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
