3 problems

Learned-bias RVFL conjecture. RVFL with learned biases outperforms RVFL without learned biases most strongly in the regime of high-dimensional balls with small radii.

Akyildirim–Teichmann conjecture. Almost surely, the signature features of $X$ can be reconstructed from its randomized signature $(Y_T^y)_{y\in\mathbb{R}^N}$.

Consider a random depth $L$ neural network with input dimension $n_0$, hidden layer widths … output dimension $n_{L+1}$ and non-linearity $\sigma$. Suppose that the network is tune…