The Big Programming Thread - Page 944

test 'should not chosen default pipeline with related columns' do

On February 08 2018 18:33 WolfintheSheep wrote:

If I recall, you've asked similar questions before, and the general theme seems to be that you're putting a lot of undue emphasis on "decluttering" your code. Clean code should come naturally no matter what you're programming, and hacking functionality together is the opposite of that.

Long-short, always close your DB adapters. There may be niche cases where you will want to leave a connection open, but you should be able to clearly say why that's necessary from a system perspective.

Also, when you say "without...wrapping it in a pointless function", I hope you're just wording that poorly, because encapsulation is a fundamental aspect of OOP.

    boolean isAuthentic(String email, char[] password) throws NoSuchAlgorithmException {
        AccountDbAdapter db = new AccountDbAdapter(getContext());
        boolean isAuthentic = db.isAuthentic(email, password);
        db.close();

        return isAuthentic;
    }

boolean isAuthentic = Session.getAccountDbAdapter().isAuthentic(email, password)

On February 10 2018 02:53 travis wrote:
Is it possible to tell from analysis of runtime and/or function calls if asymptotic complexity is polynomial, rather than exponential? By exponential I mean including 2^1.00001, for example.

Or to come to such a conclusion would that require mathematical analysis of the code ?

On February 09 2018 08:09 WarSame wrote:

Yeah, I'm trying to learn how to code cleanly through this project, so sometimes I'm definitely going over the line.

By "wrapping it" I meant pointless functions like this:

    boolean isAuthentic(String email, char[] password) throws NoSuchAlgorithmException {
        AccountDbAdapter db = new AccountDbAdapter(getContext());
        boolean isAuthentic = db.isAuthentic(email, password);
        db.close();

        return isAuthentic;
    }

I would like to replace that with a single, in-line function call, maybe like:

boolean isAuthentic = Session.getAccountDbAdapter().isAuthentic(email, password)

Since I have functions like this fairly often it would save quite a lot of unnecessary code.

On February 10 2018 03:32 WolfintheSheep wrote:

So, why do you feel that code is unnecessary?

On February 10 2018 04:28 WarSame wrote:

It seems like my code is repeating itself, making the intentions of the code less clear.

On February 10 2018 09:33 travis wrote:
So, I still have been going hard on working to find a reasonable exact solver for the TSP, and even if I haven't solved it, I have improved my results to a point where I feel they are pretty impressive.

I tested it 1000 times vs a brute force solver at n=9 and it matched the brute force all 1000 times, I tested it 100 times vs a good greedy solver I found online for n = 20 and it beat or tied (almost always beat) the greedy solver 100/100 times.

I tested it on two of the data sets here http://www.math.uwaterloo.ca/tsp/world/countries.html, n=29 and n=39. It solved both of those, in 35 seconds and 50 seconds.

For benchmarking on randomly generated matrices (which are more difficult than the typical matrices I can find online based on real data), I see times like this (some overhead probably presented by the profiler)

(these are averages, there is quite a bit of variance)
@n

11: .1 seconds (compared with the roughly ~1 minute of brute force)
12: .17 sec
13: .26 sec
14: .42 sec
15: .7 sec
16: 1.1 sec
24: 13 secs
32 55 secs

this is on my desktop pc, which is probably average in terms of speed. it's in python, if I rewrote it in C (and if i was good enough to rewrite it in C), I could probably make it go much much faster.

I of course need to test bigger datasets, but even if it messes up somewhere I am still very proud of what I have accomplished doing this stuff (my life has revolved around it for the last 3 months... ive been obsessed).

I do think I still have a couple ways in mind to make it perform faster and lose less memory (it uses a lot). Right now it will soon approach the point where my 6gb of ram is not enough.

On February 10 2018 14:36 Neshapotamus wrote:


Here are some other ways people are solving TSP.

https://github.com/DiegoVicen/ntnu-som

The basic idea is to use a self-organing map (unsupervised neural network model).

You should share your code with us and explain your approach! While building code is nice, writing and explaining your method should deepen your understanding of your algorithm. Most likely, your applying techniques that people have already thought about.

On February 09 2018 08:09 WarSame wrote:

Yeah, I'm trying to learn how to code cleanly through this project, so sometimes I'm definitely going over the line.

By "wrapping it" I meant pointless functions like this:

    boolean isAuthentic(String email, char[] password) throws NoSuchAlgorithmException {
        AccountDbAdapter db = new AccountDbAdapter(getContext());
        boolean isAuthentic = db.isAuthentic(email, password);
        db.close();

        return isAuthentic;
    }

I would like to replace that with a single, in-line function call, maybe like:

boolean isAuthentic = Session.getAccountDbAdapter().isAuthentic(email, password)

Since I have functions like this fairly often it would save quite a lot of unnecessary code.

On February 10 2018 17:16 Acrofales wrote:

If you think you're really into something new, don't share. Depending on your goals, you can write a scientific article, or try to monetize your idea.

But for it to be really good you need to be better than state-of-the-art approaches (such as those self-organizing maps), and not just better than brute force and greedy approaches.

And it's a very popular problem to work on, so the competition is strong. I wouldn't get my hopes up that you have actually made a breakthrough. But it's always a (slight) possibility. However, if you haven't actually made a breakthrough, sharing with peers is indeed a great way to improve (it's also a great way if you have made a breakthrough, but then it's more important to first protect your idea from being stolen; assuming you care about getting credit for it).