As such, attempting to study generalization by fitting models to random data is a category error, since such data offers no possibility of generalization. To understand generalization, you must look at the relationship between models and natural data manifolds.
-
-
Prikaži ovu nit
-
In the future, AI will be capable of extrapolation, not just interpolation, i.e. broad generalization, & further, general intelligence. This is an endeavor entirely orthogonal to interpolation / curve fitting, which will be achieved through an entirely different set of techniques
Prikaži ovu nit -
Of course, interpolation & pattern recognition are nevertheless a fundamental component of intelligence -- just not the most powerful. Being able to do it well -- by refining deep learning -- is immensely valuable
Prikaži ovu nit
Kraj razgovora
Novi razgovor -
-
-
I’m a fan of Roman history and I follow multiple accounts related to that. I suddenly saw Cicero talking about ML and I was like “What the f*%k”. Then I saw the name and it all made sense
.Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Would you be able to share some of the hypotheses on the types of latent manifolds we could create or induce in models? Tolman-Eichenbaum machine seems interesting?
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
dear François. the models explored as of yet are useful discrete compute modules. but a general intelligence is a class of continuous local automorphism in the space of reality. and of course, we question how to safely construct one artificially, rather than by human reproduction
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Would it be fair to say that a NN is just a way of specifying a family of functions which we use for curve fitting, and that we could also use other families, such as wavelets
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Isn't generalization already an extrapolation? After which the model can interpolate. Isn't THIS the "learning" that happens?
-
"learning" is just converting sample signals into a more efficient form... matrix factorization is "learning"...
- Još 1 odgovor
Novi razgovor -
-
Tweet je nedostupan.
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.