/comsci/ - Computer Data Science mega thread

[Anonymous No.12608644]

Top 10 programming languages according to /sci/
1) Python
2) Matlab
3) julia
4) rust
5) javascript
6) ruby
7) sql
8) c#
9) c++
10) r

[Anonymous No.12608669]

*1)c++

[Anonymous No.12608736]

1: c--

It's the less powerful, less efficient, more syntactically confusing obscure little brother of c

[Anonymous No.12608917]

>ruby
bullshit

[Anonymous No.12608935]

R

[Anonymous No.12608972]

>>12608644
How the fuck do I model something that has variable data e.g. modeling races with different number of runners?

[Anonymous No.12609006]

>>12608644
Unironically I find SQL faster and easier to use over pandas half the time, the only reason it's not higher on the list is that it's abstracted away in everything nowadays so your average code monkey doesn't know what it's capable of.

[Anonymous No.12609908]

>>12608644
>4) rust
>5) javascript
>6) ruby
>7) sql
>8) c#
>>>/g/et out

[Anonymous No.12610378]

>>12608644
WRONG!

1) Haskell
2) C
3) APL
4) R

/thread

[Anonymous No.12611941]

>>12608972
Cna you be more specific? This actually occurs quite often. For instance, you can model each runner with its own prob dist and then write the likelihood of each race as the joint. You can also create learnable embeddings for each racer, perform some operations on the embeddings, then sum them together to get a representation for a race. Its hard to tell the right way without more context.

[Anonymous No.12612253]

>>12611941
>Cna you be more specific?
So basically I have an array of attributes for each runner(age, height, weight, etc) BUT the number of runners varies, so a race can have 5 to 10 runners, so the shape of the tensor changes unless I create a separate model for each race based on number of runners

[Anonymous No.12612287]

>no C
>that image
data "science" is so cringe

[Anonymous No.12612298]

>>12608736
you're not meant to program directly in it. It is an intermediary language in GHC's pipeline.

[Anonymous No.12612775]

>>12612253
I think it depends on your task. Are you trying to predict the ranking of the runners? In this case, your problem is quite similar to that of webpage ranking in the IR field, which could be a good starting point. If your doing some general classification or regression on the entire race, you could just aggregate your embeddings for each individual runner to get an embedding for the entire race.

[Anonymous No.12612867]

>>12612775
Yeah basically modelling which runner won the race and output needs to be probability for all runners in the race, i.e. runner 1 has a 10% chance of winning, etc