[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 501 KB, 624x400, connectome.png [View same] [iqdb] [saucenao] [google]
10988120 No.10988120 [Reply] [Original]

What's the best branch prediction algorithm? Do Deep Neural Networks outperform the classic algorithms? Is there anything better than Deep Neural Networks for branch prediction?

>> No.10988385

>>10988120
>best branch prediction algorithm
You don't actually do any of this, do you? Everyone knows it depends largely on implementation. There are many, many different types of neural networks and also many, many different types of "classic algorithms." There are conditions in which each algorithm has its own optimal environment for peak performance, and a good many of the research papers of this area discuss said environments as well as comparisons to other algorithms with respect to the one(s) used.

>> No.10988473

>>10988120
If you mean classification algorithm then a random Forest training set generator paired with a strong classical classifier like k-nearest-neighbor is probably the best. I’ve seen papers where it is shown to (in general) greatly outperform neural networks with the added benefit of not being incomprehensible nonsense

>> No.10988478

>>10988473
I’ll also add that >>10988385
Is right in that it depends on the implementation. There are theorems that prove there is no inherent superiority of any classification algorithm without prior knowledge of the data.

>> No.10988827

>>10988473
No, I mean branch prediction algorithm.

>> No.10988845

>>10988120
>What's the best branch prediction algorithm?
Depends on the scenario

>Do Deep Neural Networks outperform the classic algorithms?
It depends. What are you trying to do?

>Is there anything better than Deep Neural Networks for branch prediction?
Absolutely. Why set up a computationally heavy neural network to find the support/confidence of two items when you could simply calculate this with ease?

There's no easy way out of learning about data mining. You need to know what you're asking, otherwise the question is useless.

>> No.10988855

>>10988845
I know what I'm asking. I'm asking for more information about branch prediction. Don't be so suspicious about my intentions.