Exponential Approximation Rates and Parameter Efficiency of Learnable Bernstein Activations

Ibrahim Albool, Malak Gamal El-Din, Salma Elmalaki, Yasser Shoukry

arXivarxiv/2602.04264Feb 4, 2026cs.LGscore 0

The choice of activation function fundamentally shapes the representational capacity and parameter efficiency of deep neural networks, yet most widely used activations lack rigorous theoretical guarantees on these properties. We provide a theoretical analysis of DeepBern-Nets (DBNs) -- networks employing learnable Bernstein polynomial activations -- showing that their approximation error decays with the network depth $L$ and the polynomial order $n$ with a rate of $\mathcal{O}(n^{-L})$, exponentially faster than the polynomial rate of ReLU architectures while remaining fully differentiable.…

SourceJSON

Comments and reviews

No comments yet.

API

curl /api/v1/records/arxiv/2602.04264
curl -X POST /api/v1/records/arxiv/2602.04264/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2602.04264/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"body": "…", "review": true}'