so… let’s play a game !!!
i’ll show you a series of patterns, and each pattern maps to a value of either 0 or 1
now consider this pattern, which number would this pattern map to?
if you said 0 you’re probabily right
if you said 1 you’re also probabily right
¯\_(ツ)_/¯
the point here is that whatever your guess was…
- you had no idea what each of the tiles represented
- you had no idea what the co-relation between the tiles was
- you had no idea what 0 or 1 represents
just like AI and this guy here
he has 0 understanding of what the puzzle means and what his guess means, but he made the correct guess
The Learning Problem
the game we just played right now, is presicely what every ML/DL/RL/AI algorithm does
- take data that has been cleaned/processed (into a puzzle)
- find a pattern in the data
- use the pattern on new data (it has never seen before)
- hope it gets appriciated
but keep in mind that not every problem can be solved this way
because…
- not every problem has (enough) data on it
- not every data has a pattern in it
some of them also/already have a (not so) simple mathematical answer to them, you don’t need rely to finding patterns
but for the rest, they can probabily (maybe not efficiently) be solved by having a computer find a pattern AKA learning from data
How does a Computer Learn from Data
short answer, you solve the learning problem 😝
long answer to solve any learning problem (ML, DL, RL)👇
- unknown solution: think of this the true pattern/source of all the data we have
- training data: data that we have collected and processed so a computer can find a pattern
- learning algorithm: the actual algorithm that refines a model so the model learns from the data
eg: Perceptron Learning Algorithm, Backpropagation Algorithm, etc
- hypothesis set: a set of all possible models that can find a solution
eg: Perceptron, Neural Networks, etc
- final hypothesis solution: a pattern with the heighest accuracy
while this IS the long answer to solve any learning problem
this long answer lacks details… the details which would be dealt with in the forthcoming blog articles
untill next time !️ ✌
or you could spot me in the wild 🤭 i mean instagram, twitter, linkedin and maybe even youtube where i excalidraw those diagrams