We don’t have any other access to reality but counting the noisy bits on our systemic interface. Evolving priors is part of counting. Higher order relationships between observed patters is still counting.
-
-
Replying to @Plinz
Josha, again I don't disagree. I am not even against correlation. Just not frequency based correlations of non universal & non grounded knowledge. Speaking of noise, I agree. However, of one has no high abstraction of grounded knowledge, one would not even be aware of noise.
1 reply 0 retweets 0 likes -
Replying to @eyad_nawar
What matters is not what opinion you hold, but what arguments support it. I don’t understand your prerequisite of “universal, grounded knowledge”. Are you sure your epistemology is sound?
1 reply 0 retweets 0 likes -
Replying to @Plinz
I do show that in my next article, where I talk about how LeCun's continuous functions are not actually continuous, (they are, mathematically) but not cognitively. Because the function's domain is limited by the range of values that the probability distribution of the data take.
1 reply 0 retweets 1 like -
Replying to @eyad_nawar @Plinz
Let me quote William Briggs: "Mathematical equations are lifeless creatures; they do not “come alive” until they are interpreted, so that probability cannot be an equation. It is a matter of our understanding. […] Mathematical statistics does not withstand interpretation of the
1 reply 0 retweets 0 likes -
Replying to @eyad_nawar @Plinz
values it estimate." DL models (as any statistical model) do not understand what values they predict. Its a lifless gradient based optimization of a loss function. That is why, values outside of the distribution are not "interpreted" into their actual classes, even though the
1 reply 0 retweets 0 likes -
Replying to @eyad_nawar @Plinz
belong to the same domain. That is mathematically & statistically why no generalization takes place. Grounded deterministic knowledge (the knowledge we use to efficiently talk about the relation among things (which you mentioned)) is a way to interpret highly variant inputs.
1 reply 0 retweets 0 likes -
Replying to @eyad_nawar @Plinz
If it has wheels, has a body, changes your external state from one sub-environment to another, belong outdoors, + etc then it is probably a car, even if you never seen any like it before.
2 replies 0 retweets 0 likes -
Replying to @eyad_nawar @Plinz
I believe that all stochastic phenomenon ground to a deterministic model of the universe, or at least how we perceive the universe. Else we can't interpret stochastic events I apologize for messing up your timeline, we have different views. I still look up to you though. Cheers!
1 reply 0 retweets 0 likes -
Replying to @eyad_nawar
I don't think that we are on the same page. I don't think that human minds learn of a reality beyond the functions that they approximate, and there is eventually no fundamental difference between statistical, causal and econometric approaches.
1 reply 0 retweets 1 like
Deterministic functions can be understood as a subclass of statistical models, preferable because they yield the same result every time, so they have short description lengths for observed events, and deliver tight predictions of as yet unobserved events.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.