Northrop Grumman Inquiry

[Madeleine Gagné]

Hello All,

I hope this notice finds you well. My name is Mac Gagne and I'm a current employee of Northrop Grumman's Future Concept's team. One of our currenct projects is currently utilizing your NetworkX Dijkstra algorithm in python to aid in operational work pertaining to satellite communications. As our team always likes to stay on top of developments in the space and modeling spheres, we've recently learned about developments in the gradient descent algorithm (as highlighted in the following Quanta Magazine article). From our understanding of this news, the gradient descent algorithm (which, if we're not mistaken, your Dijkstra algorithm employs) has been found to cover a number of edge cases not previously considered. As such, we wanted to reach out and ask your team if these recent developments and changes to the gradient descent algorithm have been incorporated into the Dijkstra algorithm. Our team wants to make sure we're using the most developed form of gradient descent, and we'd love to know if these recent developments in the scientific field have had an impact on the great work NetworkX is advancing. Any information, guidance, or advice on this topic would be much appreciated.

Northrop email: Madelei...@ngc.com 

(of feel free to respond to this message!)

Thank you so much for your time and all my best,

~Mac Gagne

[Dan Schult]

Hello Mac Gagne,

In my understanding, Dijkstra's algorithm doesn't involve using gradient descent. Gradient descent moves from one candidate solution to a (hopefully better) new candidate solution based on the gradient of the function to be optimized. Dijkstra constructs a solution by carefully considering edges. So, in Dijkstra we aren't iteratively improving a solution.

We do try to keep up on the latest developments for algorithms and implement them when we can.  Help with doing that is always appreciated. 

Best,

Dan Schult

[Madeleine Gagné]

Heya Dan!

Thanks so much for this further clarification. I really appreciate it! Your recommendation has really pointed us in the right direction.

All my best and thank you so much for your time!

~Mac