Please enable JavaScript.
Coggle requires JavaScript to display documents.
Neural Network: At a Glance - Coggle Diagram
Neural Network:
At a Glance
what is a neuron
inspired by
human brains
ability to learn intuitively
what is it
a network of functions
input (dandrites)
process (weight & bias)
output (activation function)
why is it important
good for discovering hidden patterns
weights & bias
what is it
the only thing that can be tuned
weight
similar to gradient
"how important is this input"
"how much should this input influence the output"
bias
constant value (offset)
similar to the 'y intercept'
input * x + y
who changes weight & bias
the deep learning will figure out themselves
not manually tuned.
backpropagation
introduction to ANN
what is ANN
deep learning's core foundation
network of neurons, inspired by human brains
why ANN is important
very great at classification & regression
the future
DL vs. ML
feature engineering
[ML] intensive feature engineering
[DL] light feature engineering
data size
[ML] doesn't need big data
[DL] needs a big data
activation function
what is it
function that determines should it "fire"
basically the output
normalized (1 to 0) or (1 to -1)
why is it important
transferring signal to the next neuron
normalizes the output
weight could be more than 1 or less than 1
so is with the bias
activation function lists
linear activation functions
(not recommended)
non-linear activation functions
logistic function
tanh function
ReLU function
ANN architectures