cambridgeuniversitypress / firstcoursenetworkscience Goto Github PK

Not an error, but just notice that reported Python code does not always follow common naming conventions on variables: Python's variable are expected to start with small case characters, and names are exptected to be not too general.
Using one capital letter to name a variable is generally considered too naive, as reported in many tutorials, such as: https://medium.com/@dasagrivamanu/python-naming-conventions-the-10-points-you-should-know-149a9aa9f8c7

There is a reason behind such a convention: expert programmers want to avoid ambiguity with classes; in fact, capitalized names are usually given to classes.

For example,
G = nx.Graph()
could be changed to something like:
graph1 = nx.Graph()

(also note that the name of the class 'Graph' is capitalized).

I am aware that such ambiguity has been introduced in the networkx official tutorial in the first place, but students with some knownledge on Python may find reported examples odd-looking.

In Section 6.3.2 Modularity Optimization, shaded box illustrating the Louvain algorithm, end of item 2: "...with weight equal to the number of internal links." -> "...with weight equal to twice the number of internal links."

Should the code in this repo be CC-BY or something like that?

Also, who owns the copyright for this code?

If nodes represent actors, as said in the caption and in the inline text, it does not make any sense to color nodes according the film genre, am I right? an actor may play different roles in different movies of differerent genres. Moreover colors match suspiciously too well with clusters: it looks like that colors have been assigned after the membership to a given modularity class. Just a guess.

In the ch 7 tutorial, there is a deprecation warning after the 1st cell. @clayadavis should we remove it?

If so, add the 'W' label in the table in chapter 1.

Is it possible to change the network (eg, rewire edges) in state_transition()? One would need to do that to implement the model in section 7.3.3 of the book...

We need to proofread all tutorials, as we have found several typos.

• Appendix
• Ch 1
• Ch 2
• Ch 3
• Ch 4
• Ch 5
• Ch 6
• Ch 7

Wait for feedback from editor as to whether to only post solutions to selected problems (on Github), or to all exercises (only accessible to instructors), or a combination of both

Apparently there are no arrows in the picture.

Should we require a minimum version of NetworkX in the tutorials?

One exercise requires greedy_modularity_communities, which became available in v.2.2.

Another exercise requires the version of asyn_lpa_communities that we fixed, in the latest version (2.4).

#19

In Eq. 2.2, 2.3, and 2.5 the sum over the pairs of nodes is supposed to include each pair (i,j) once or both (i,j) and (j,i) depending on whether the network is directed or undirected. This is mentioned in the text but it may be unclear and confusing. We could rewrite the formulas so that the notation $\sum_{i,j}$ always means the same thing.

Thanks to @giaruffo for flagging this issue!

I could be wrong, but co-citing papers (connected with dashed blue links to the cited paper) should be both on the left of the target paper, if links points always backward in time.

• Post solutions to selected exercises (#6)
• Add link to CUP website (https://www.cambridge.org/academic/subjects/physics/statistical-physics/first-course-network-science?format=HB#resources)
• Reorganize repo

Santo points out that nx.read_edgelist producing str-typed node labels is non-obvious. This could be improved by using nodetype=int in the Enron example and including an aside explaining why.