🧵 Markov models
Raphael at Tue, 10 Sep 2024 01:02:25 UTC No. 16370864
If this uses probability theory then why doesn’t it follow the normal distribution?
>t. 100 FSIQ anti memer
Anonymous at Tue, 10 Sep 2024 01:19:49 UTC No. 16370887
>>16370864
It does. It's just being binned to make the math easier.
Raphael at Tue, 10 Sep 2024 01:21:10 UTC No. 16370890
>>16370887
Proof
Anonymous at Tue, 10 Sep 2024 01:29:16 UTC No. 16370896
>>16370890
Look at the sunny state for example. 0.7 probability of staying sunny and 0.3 probability of turning cloudy. The normal distribution is being binned at prob = 0.7.
[0.0,0.7) -> Sunny bin
[0.7,1.0] -> Cloudy bin
raphael at Tue, 10 Sep 2024 02:58:09 UTC No. 16370989
>>16370896
so theyre pseudo neurons with probabilities pre programmed and it predicts an outcome from state changes
Anonymous at Tue, 10 Sep 2024 03:02:36 UTC No. 16370994
>>16370989
>probabilities pre programmed
... according to a binned normal distribution.
Exactly. I think you get it. Any more questions?
raphael at Tue, 10 Sep 2024 03:04:04 UTC No. 16370996
>>16370994
one more anon, what is a binned distribution?
Anonymous at Tue, 10 Sep 2024 04:16:43 UTC No. 16371084
>>16370996
When you break up a range into several pieces or subranges. For example, start with the range [0,1] and break it into three pieces, or bins, like [0.00,0.33), [0.33, 0.67), and [0.67, 1.00].
raphael at Tue, 10 Sep 2024 04:21:43 UTC No. 16371095
>>16371084
then why do markov models not follow a binomial distribution
Anonymous at Tue, 10 Sep 2024 04:26:23 UTC No. 16371104
>>16371095
You can pick any distribution to bin. The normal distribution is just traditional.
Raphael at Tue, 10 Sep 2024 05:00:47 UTC No. 16371173
>>16371104
Can you show me one
Anonymous at Tue, 10 Sep 2024 12:33:47 UTC No. 16372060
>>16371173
Youve seen two now.
Anonymous at Tue, 10 Sep 2024 14:08:48 UTC No. 16372233
Anonymous at Tue, 10 Sep 2024 15:12:24 UTC No. 16372321
>>16370996
Google can help you with this.
https://www.allaboutcircuits.com/te
raphael at Tue, 10 Sep 2024 19:27:51 UTC No. 16372911
>>16372233
thats the normal distribution retard
raphael at Tue, 10 Sep 2024 19:28:52 UTC No. 16372912
nvm i got it thanks anons
raphael at Tue, 10 Sep 2024 19:38:03 UTC No. 16372927
>>16372321
why didnt you just say it was a bar graph distribution nigga
Anonymous at Tue, 10 Sep 2024 19:39:44 UTC No. 16372930
>>16370996
>what is a binned distribution?
>>16372911
a binned normal distribution, you retard.
Anonymous at Tue, 10 Sep 2024 19:42:10 UTC No. 16372939
>>16372927
because a bargraph is a display tool for visualizing binned distributions, homeslice
Anonymous at Tue, 10 Sep 2024 19:45:49 UTC No. 16372946
raphael at Tue, 10 Sep 2024 20:01:47 UTC No. 16372998
thanks anons
RL uses markov models which is weird
Anonymous at Tue, 10 Sep 2024 20:04:53 UTC No. 16373002
>>16370864
Markov models are only useful for selling financial products and have no predictive value in real life.
Anonymous at Tue, 10 Sep 2024 20:25:58 UTC No. 16373057
>>16370864
Can you tell me one thing besides coin flip, that does follow normal distribution?
Anonymous at Tue, 10 Sep 2024 20:40:04 UTC No. 16373079
>>16373057
Size
Anonymous at Tue, 10 Sep 2024 20:43:41 UTC No. 16373085
>>16373002
That's only because most people just guesstimate the model parameters in Markov models. Compare to LLM where no one even cares what the parameters are as long as we tell the investors that we have 17.25 trillion of them.
Anonymous at Tue, 10 Sep 2024 21:02:59 UTC No. 16373129
>>16373085
Lol. LLM is essentially large language markov. A screenshot that pretends to be timeless.
Anonymous at Tue, 10 Sep 2024 21:10:46 UTC No. 16373151
>>16373079
Size of anything human is skewed.
Anonymous at Tue, 10 Sep 2024 21:15:05 UTC No. 16373162
>>16373151
I'm sorry your dick is so unhumanly small.
Anonymous at Tue, 10 Sep 2024 21:31:56 UTC No. 16373205
>>16373151
Anthrocentrist. That's why you normalize the data.
Anonymous at Tue, 10 Sep 2024 21:33:35 UTC No. 16373211
>>16373129
Either way, 70% chance of sun tomorrow.
Anonymous at Tue, 10 Sep 2024 22:27:54 UTC No. 16373307
>>16373211
In a covering set that no one cares about.
Anonymous at Tue, 10 Sep 2024 22:49:35 UTC No. 16373340
>>16373307
>no one cares about.
80% of people disagree with you.
Anonymous at Tue, 10 Sep 2024 22:53:43 UTC No. 16373351
>>16373340
No one cares about 70% chance of sun on earth if they know 70% earth is water.
Anonymous at Tue, 10 Sep 2024 23:01:53 UTC No. 16373368
>>16373351
80% of people disagree with you.
Anonymous at Tue, 10 Sep 2024 23:07:32 UTC No. 16373372
>>16373368
100% of people disagree with you.
Anonymous at Tue, 10 Sep 2024 23:21:11 UTC No. 16373379
>>16373372
You got one model and you can't even read it correctly. Sad. Many such cases.
Anonymous at Tue, 10 Sep 2024 23:29:51 UTC No. 16373387
>>16373379
So far I have two and you have one
Raphael at Wed, 11 Sep 2024 00:41:35 UTC No. 16373496
>>16373129
I don’t think you understand how LLMs work
raphael at Wed, 11 Sep 2024 00:56:49 UTC No. 16373516
>>16373496
this nigger is so retarded he even doesnt know how abstraction works
Anonymous at Wed, 11 Sep 2024 01:07:32 UTC No. 16373523
>>16373496
we used to call it gigo
raphael at Wed, 11 Sep 2024 02:20:14 UTC No. 16373563
>>16373523
still not a markov model
Anonymous at Wed, 11 Sep 2024 02:24:19 UTC No. 16373567
>>16373563
every markov model is gigo