a2liu / dsfp Goto Github PK

Write new add methods in each type of network

_{Sent with GitHawk}

Nodes don’t need to contain user objects, they only need to hold references to the row of the user in the MySQL database. They may then also have to hold a few more bits of info, like weight and list of ratings. May have to create an object to hold this info and rework the “vertex” class.

_{Sent with GitHawk}

Addvertex should return the name of the vertex it just created.

_{Sent with GitHawk}

Write more comments

_{Sent with GitHawk}

_{Sent with GitHawk}

that way it's easier to implement features that have to do with standard deviation later

Each node in the ratings graph should have a collection of edges going from it to other nodes that have rated it, instead of the other way around. The weight of the edges will still be a rating, but it'll be a rating that's received instead of given. Lists of ratings for the user can be generated through SQL queries instead of through the network; the network is built for searching for friends and for updating the ratings table.

Need to use rounding in order to prevent a single update from traversing the entire network.

_{Sent with GitHawk}

Util Package. Will definitely be useful in future work.

May be able to edit algorithm to make ratingsNode objects mirror the behavior of neurons. Probably ineffective to in a language like Java, but basic concept can be ported over to Python later if need be.

Also, might be able to change algorithm to accept other parameters, i.e. tags, and then use modified algorithm to do stuff with Uber driver ratings or Yelp ratings (businesses rating businesses). Or even scientific publications (reputable and irreputable journals that reference each other). Still needs to be sorted out though.

Users should start with a rating of 1. As they get ratings, their initial ratings get replaced by real world ratings. Also, people that have been rated more, and rated more consistently, also have higher rating weights.

Will probably need to make an “Updater” class to update the network. Otherwise handling the references might make the code too messy and hard to read

_{Sent with GitHawk}

That is all.

_{Sent with GitHawk}

_{Sent with GitHawk}

Ratings may cause feedback loop in network. To fix, each node should have a counter that checks for how many times it has been updated, and slowly increase the amount of difference in rating necessary to cause an update. I think threshold should probably increase quadratically. May have to create another network class that extends NodeNetwork to implement updates to the network.

_{Sent with GitHawk}

Will need to store an image of network before update chain and find difference between the old and new network, and send the difference back to the database.

_{Sent with GitHawk}

Add a secondary hashtable that contains previous iteration's information first, then update nodes off of that info so that updates are all effectively simultaneous.

ALSO THIS WAY TRAVERSAL ORDER DOESNT MATTER AND THAT IS GOOD.

There needs to be an easier way to load edges into the network than just filling a csv using google sheets/excel. Maybe I could write more functions, but I'd prefer not to. It's annoying coding in 2-3 different languages simultaneously. I'd much rather just code in one (Java). Also, I dislike coding in Google sheets; it takes whole minutes to process simple calculations over a few hundred thousand cells, and has a 2 million cell limit.

UI could be done with JavaFX Control? Or just normal JavaFX

It's technically better than MySQL.
https://www.microsoft.com/en-us/sql-server/sql-server-downloads

Just keep the rest of the network static. No need to keep whole network exact, calculate exact values for specific subsection then let the rest go.

_{Sent with GitHawk}

The self rating shouldnt stay at 1. It makes attaining a 5 practically impossible, and also makes it harder to reach equilibrium.

_{Sent with GitHawk}

Dont return edge or edgepair objects from network or graph public methods

_{Sent with GitHawk}

Add config file support for server package
Add user file support for user class
Add admins with extra priveledges
Add datastructure for ratings system
Add modular ID confirmation/security features
Add tons of other stuff

_{Sent with GitHawk}

Idk how to implement messages yet but i can probably try? I guess it’d just be another edge network with repeat edges and each edge holds a message, the order of the primary key determines message order, and the messages are dumped into the DB w/ synchronized methods. There might be some probs with retrieving though. Don’t wanna parse through entire edge list to receive a single message.

_{Sent with GitHawk}

See above.

_{Sent with GitHawk}

Remove strings from code that directly reference stuff in computer, put it in a property file instead and add that file to .gitignore

_{Sent with GitHawk}

Include a parameterized type T and use that to set the name/key of each vertex to make it easier to use toString methods

_{Sent with GitHawk}