davinci-net.
Share Brilliantly
optimization
Gradient Descent - Constant Step Sizes
Overview
Intuition
{% \vec{ \theta_{k+1} } = \vec{ \theta_k } - \nu_k \vec{g_k} %}
where {% \nu_k %} is chosen such that {% || \nu_k \vec{g_k} || %} is some constant.
Contents