What to feed to the input of the neural network? Your ideas... - page 65

New comment
 
Forester #:
Not clear. Describe the evaluation algorithm.
In a multiplication table from 1 to 10, there are 100 choices of combinations. So, 100 correct answers to all the multiplication variants of numbers are required for complete memorisation.
 
Andrey Dik #:
In the multiplication table from 1 to 10 there are 100 variants of combinations. So, 100 correct answers to all multiplication variants of numbers are required for complete memorisation.
That is, going through the different variants, as I suggested earlier?
 
Forester #:
You mean iterate through different options as I suggested earlier?

maybe I didn't understand you right away, what are we talking about? do you have a way to avoid iterations and get a 100% memorisation result right away?
 

There is no separate metric for the quality of "memorisation" in MO. Memorisation and generalisation are two related elements of the same process, two sides of the same coin, which share a common metric such as logloss.

Valuable experts could finally read the theory of MO.

 
Andrey Dik #:

Maybe I didn't understand you right away, what are we talking about? Do you have a way to avoid integrations and get a 100% memorisation result right away?
I just don't agree with the opinion that a record becomes memorisation only after evaluation.
When writing data to a database (or to a tree, or to a maximal learning clustering model), it just writes it as it is, without any extra actions. The database has no idea what it is given to memorise (multiplication table, Ohm's law, another formula or market near-random). What it is given to remember/record, it stores in database rows, leaves, clusters.
 
After training, or rather after training, we can decompose the model into generalisation and memory components.
 
Ivan Butko #:

Before talking about an applied kind of learning (algorithm), and even learning theory, I wanted to draw attention to the full picture of the local sabj (learning). From and to.

Based on some answers,
Learning is agradual mastering of a way of interacting with the world, where learning becomes not so much a process of accumulating knowledge as a constant adaptation, understanding of patterns and the ability to apply them in practice in a new context.

Learning is the accumulation of knowledge or experience, Maestro :)))
 
A counter question to the experts: what is the difference between knowledge acquisition and memorisation?
 
Forester #:
I just disagree with the opinion that recording becomes memorisation only after evaluation.
When writing data to a database (or to a tree, or to a maximal learning clustering model), it just writes it as it is, without any extra actions. The database has no idea what it is given to memorise (multiplication table, Ohm's law, another formula or market near-random). What it is given to remember/record, it stores in database rows, leaves, clusters.

you are talking about recording, not memorisation. memorisation is a fundamentally different process, recording cannot become memorisation in any way.
 
Maxim Dmitrievsky #:
A counter question to experts: what is the difference between knowledge acquisition and memorisation?
In order to answer this question, they first need not to confuse a database and a knowledge base... But apparently not in this life
1...58596061626364656667
New comment