5 problems

0 votes0 replies1 view
Depth-three stabilization conjecture for univariate ReLU variation balls
Let ${\mathcal{B}}_L$ denote the unit ball at depth $L$ in the univariate ReLU setting. Depth-three stabilization conjecture. The unit balls stabilize after depth $L=3$ , namely … C…
Conjectureopenfunction-spaces neural-networks reluarXiv posted 2 days ago
0 votes0 replies0 views
The optimal lower Lipschitz bound conjecture for ReLU layers
Optimal lower Lipschitz bound conjecture. It was conjectured that
Conjectureopenlipschitz-continuity neural-networks reluarXiv posted 2 days ago
0 votes0 replies0 views
Completeness conjecture for incremental quadratic constraints of ReLU
Completeness conjecture. We initially conjectured that
Conjectureopenneural-networks quadratic-constraints reluarXiv posted 2 days ago
0 votes0 replies1 view
Zero-risk convergence conjecture for random-initialized ReLU gradient descent
Consider the plain vanilla gradient descent optimization method with random initializations for artificial neural networks with one hidden layer and ReLU activation. Let the networ…
Conjectureopengradient-descent neural-networks reluarXiv posted 2 days ago
0 votes0 replies0 views
Universality of capacity for multilayer networks with ReLU hidden layers
A multilayer network has ReLU transfer functions in its hidden layers, meaning that each hidden-layer unit applies the function $t\mapsto\max(0,t)$ . Capacity universality conjectur…
Conjectureopencomputational-capacity neural-networks reluarXiv posted 2 days ago

5 problems

Depth-three stabilization conjecture for univariate ReLU variation balls

The optimal lower Lipschitz bound conjecture for ReLU layers

Completeness conjecture for incremental quadratic constraints of ReLU

Zero-risk convergence conjecture for random-initialized ReLU gradient descent

Universality of capacity for multilayer networks with ReLU hidden layers