Math Seminar
Artificial intelligence is similar to machine learning, but artificial learning goes far deeper. A
self-driving car is one example of artificial intelligence, and it currently has some issues. One of
them is that we are preparing machines for pre-learning. There is an issue with that. Humans
know how to drive, but we don't understand how.
Since there are a lot of unexpected things going on, humans are good at finding them quickly and
working out a solution. Computers, on the other hand, are confined to pre-learned programs and,
due to a lot of noise, are not so good at making fast unpredictable decisions. As we all know,
mathematics is full of problems that must be solved with a specific piece of knowledge, and it
can also be very unpredictable. The extent of a computer performing tasks is the extension of
pre-learned boundaries, implying that machines are restricted to their internal program.
Machines need training data to perform a new task, which means that if we give the machine
something that differs slightly from the training data, the machine will not be able to answer
those questions. In mathematical terms, we can use cross-validation to assist our machines in
distinguishing between right and wrong. Does this imply that the computer is capable of deep
learning? The response is possible. If the computer is capable of obtaining the correct response
from raw data, we may consider deep learning, but only in that domain.
Deep learning has become popular in recent years. Deep learning is a large amount of
computational data that computers have learned to use and solve new problems, regardless of the
problem. One example is an automated chess-playing system that defeated the world's best
players. Deep learning machines seem to be the future of our culture. This is due to the massive
amount of data we provided (humans).
Linear algebra plays an important role in deep learning. With today's infinite computing
resources, machine learning is expanding faster than ever. Deep learning is simply a parameter
variable function with several different variables that all have the same composition (function
composition), and using activation functions means using a non-linear function and then adding a
linear function to it. Any possible function can be correctly described with the composition of
function and network, but that is just one layer of the network, and we still don't know how this
