I read in my notes:
if we use Dijkstra $|V|$ times ($|V|$ number of vertexes) for finding all pairs shortest paths in graph $G$, We get time complexity for Dijkstra algorithm as $O(VE+ V^2 log V)$ and if we run bellman ford algorithm $|V|$ times we get time $O(V^2E)$.
The above details is not important.
I read too, in sparse graph Dijkstra works better (i.e: $O(VE+ V^2 log V)$ is better asymptotic respect to $O(V^2E)$ in sparse graph).
I think sparse graph is $ |E| \in O(V)$ or $E=o(V^2/\log V)$. Infact I have two misunderestanding:
$1)$ which one is used to show a graph is sparse commonly?
$2)$ according to $(1)$ how we can intuitively understand $O(VE+ V^2 log V)$ is better asymptotic respect to $O(V^2E)$ at least I think the reverse is true.