If venue feature vectors contains features f_0..f_n, and user interacted with venues v_0, .. v_k, define a math model that we can use to calculate user feature vectors from these facts. User's personal information should influence this as well, so if user properties are p_0, .. p_t, we calculate user feature vector
<u0,..um> = f(v_0,..v_k) * g(p_0, .. p_t)
If venue feature vectors contains features f_0..f_n, and user interacted with venues v_0, .. v_k, define a math model that we can use to calculate user feature vectors from these facts. User's personal information should influence this as well, so if user properties are p_0, .. p_t, we calculate user feature vector <u0,..um> = f(v_0,..v_k) * g(p_0, .. p_t)