Final Peer Review - hco23

[henry-ogworonjo]

Machine learning/AI has made immense progress and is attracting a lot of attention. The authors in this paper decide to investigate how the esteemed field of machine learning/AI perform relative to humans in problems that are simple to solve. They investigate the ability of deep learning algorithms to generalize on combinatorial optimization problems by focusing on the minimum spanning tree which have an expected linear time classical solution. Several approaches were considered in solving the minimum spanning tree problem. A key part of the approach is the adaptation of a convolutional neural network to learn the implicit structure of the minimum spanning tress problem.

The idea sought in the project is an interesting one as it makes sense to see how well AI does with simple task. The idea to focus on the simplest task makes good research sense. The realization and proposal of a future work is commendable. Also, the survey of related works in this field is robust.

The authors could have done a better job organizing the report such that it has the flow that one expects from a research work. The section discussing experiments is inadequate and could have been improved upon. Notwithstanding, it’s a great and interesting effort.