• Stars
    star
    124
  • Rank 278,974 (Top 6 %)
  • Language
  • Created over 5 years ago
  • Updated over 2 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Korean translation of the Keras documentation.

Korean translation of the Keras documentation

This is the repository for the Korean-language .md sources files of keras.io.

Existing files in sources/ should be edited in-line.


์ผ€๋ผ์Šค ๊ณต์‹ ๋ฌธ์„œ ํ•œ๊ตญ์–ดํŒ

์ผ€๋ผ์Šค ๊ณต์‹ ๋ฌธ์„œ์˜ ํ•œ๊ตญ์–ดํŒ์ž…๋‹ˆ๋‹ค. ์ด๋ฏธ ๋”ฅ๋Ÿฌ๋‹์— ์ต์ˆ™ํ•œ ์—ฐ๊ตฌ์ž ๋ฐ ๊ฐœ๋ฐœ์ž ์™ธ์—๋„ ์ฒ˜์Œ ๋”ฅ๋Ÿฌ๋‹์„ ์ ‘ํ•˜๋Š” ์‚ฌ์šฉ์ž๋“ค์ด ์ตœ๋Œ€ํ•œ ์‰ฝ๊ฒŒ ์ดํ•ดํ•˜๊ณ  ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ๊ทธ ์˜๋ฏธ์™€ ์šฉ๋ฒ•, ์šฉ๋ก€๊ฐ€ ์ •ํ™•ํ•˜๊ณ  ๋ช…๋ฃŒํ•˜๊ฒŒ ๊ทธ๋ฆฌ๊ณ  ์ตœ๋Œ€ํ•œ ์ž์—ฐ์Šค๋Ÿฌ์šด ๋ฌธ์žฅ์œผ๋กœ ๋‚˜ํƒ€๋‚˜๋„๋ก ์ž‘์„ฑ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ๐Ÿ“–โœ๏ธ๐ŸŒ

๋ฒˆ์—ญ ๊ฐ€์ด๋“œ๋ผ์ธ

  • ๋ชจ๋“  ๋ฒˆ์—ญ๋ฌธ์€ ํ•œ๊ตญ์–ด ์ •์„œ๋ฒ•์„ ์ค€์ˆ˜ํ•ฉ๋‹ˆ๋‹ค.
  • ๋ฒˆ์—ญ์€ ๋ฌธ์„œํ™” ๋‚ด์— ์žˆ๋Š” ๋ณธ๋ฌธ ๋‚ด์šฉ๊ณผ ์ฝ”๋“œ ์ฃผ์„๋“ค์„ ๋Œ€์ƒ์œผ๋กœ ํ•ฉ๋‹ˆ๋‹ค.
  • ๋ฒˆ์—ญ์‹œ ๋ฌธ์žฅ ๋์— ๋ถ™๋Š” ๊ฒฉ์‹์ฒด๋Š” '-ใ…‚๋‹ˆ๋‹ค'์ฒด๋ฅผ ๋”ฐ๋ฅด๋ฉฐ ๋น„์†์–ด๋‚˜ ๋ฐ˜๋ง์€ ์“ฐ์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
  • ํฐ ๋”ฐ์˜ดํ‘œ๋‚˜ ์ž‘์€ ๋”ฐ์˜ดํ‘œ๋Š”('๏ผŒ") ํŠน์ˆ˜๋ฌธ์ž๋ฅผ ์‚ฌ์šฉํ•˜์ง€ ์•Š๊ณ  ๊ธฐ๋ณธ์ ์œผ๋กœ ์ œ๊ณต๋œ ๊ฒƒ์„ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.
  • ์ฝ”๋“œ ๊ฐ•์กฐ(syntax highlight) ๋’ค์— ์กฐ์‚ฌ๊ฐ€ ๋ถ™๋Š” ๊ฒฝ์šฐ, ๊ณต๋ฐฑ์„ ๋„ฃ์ง€ ์•Š์Šต๋‹ˆ๋‹ค(e.g. model.fit()์„ ์‹คํ–‰ํ•˜๋ฉด).
  • ํ‚ค์›Œ๋“œ๋ฅผ ๋ฒˆ์—ญํ•  ๋•Œ ์•„๋ž˜์— ์žˆ๋Š” ์ž‘์„ฑ ๊ทœ์น™ ๋ฐ ์šฉ์–ด ํ†ต์ผ์•ˆ์„ ์ตœ์šฐ์„ ์œผ๋กœ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค.
  • ๊ณผํ•œ ๋ณต๋ฌธ์˜ ๊ฒฝ์šฐ ๋‹จ๋ฌธ์œผ๋กœ ๋‚˜๋ˆ„์–ด์„œ ์”๋‹ˆ๋‹ค.
  • ์›๋ฌธ ๋‚ด์šฉ์ด ๋ถˆ์ถฉ๋ถ„ํ•œ ๊ฒฝ์šฐ ์›๋ฌธ์ด ์ „๋‹ฌํ•˜๊ณ ์ž ํ•˜๋Š” ๋‚ด์šฉ์„ ์ถฉ์‹คํžˆ ์ „๋‹ฌํ•˜๋Š” ๋ฒ”์œ„ ๋‚ด์—์„œ ์‚ฌ์šฉ์ž๊ฐ€ ์ดํ•ดํ•  ์ˆ˜ ์žˆ๋„๋ก ๊ฐ„๋žตํ•œ ์„ค๋ช…์„ ๋ณด์ถฉํ•ฉ๋‹ˆ๋‹ค.
  • ๋ฒˆ์—ญ์€ ๋‹ค๋ฅธ ์–ธ์–ด๋กœ ๋œ ๋ฌธ์„œ์˜ ์˜๋ฏธ๋ฅผ ์ดํ•ดํ•˜๊ณ  ํ•œ๊ตญ์–ด๋กœ ๋‹ค์‹œ ํ‘œํ˜„ํ•˜๋Š” ๊ฒƒ์ด๋‹ˆ ๋ฒˆ์—ญ์ฒด๋Š” ์ž์ œํ•ด ์ฃผ์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค(์šฐ๋ฆฌ๋Š” ํ•œ๋‹ค ๋ฒˆ์—ญ์„).

์ž‘์„ฑ ๊ทœ์น™

  • ์šฉ์–ด ๋ฒˆ์—ญ์˜ ๊ฒฝ์šฐ ๋ฌธ์„œ ๋‚ด์—์„œ ์ฒ˜์Œ ๋‚˜์˜จ ๊ฒฝ์šฐ์— ํ•œํ•ด subscript๋กœ ์›์–ด๋ฅผ ๋ณ‘ํ–‰ํ‘œ๊ธฐํ•ฉ๋‹ˆ๋‹ค. (์˜ˆ: ์ธตlayer)
    • ๋ฐœ์Œ๋งŒ ํ•œ๊ธ€๋กœ ์˜ฎ๊ธด ๊ฒฝ์šฐ subscript๋Š” ์ƒ๋žตํ•ฉ๋‹ˆ๋‹ค. (์˜ˆ: ์ŠคํŠธ๋ผ์ด๋“œ)
    • ํŠน์ˆ˜ํ•œ ๊ฒฝ์šฐ๋ฅผ ์ œ์™ธํ•˜๋ฉด subscript๋Š” ์†Œ๋ฌธ์ž๋กœ ์ž‘์„ฑํ•ฉ๋‹ˆ๋‹ค. (ํŠน์ˆ˜ํ•œ ๊ฒฝ์šฐ: 1. ๋Œ€๋ฌธ์ž ๊ณ ์œ ๋ช…์‚ฌ ๋ฐ ๋Œ€๋ฌธ์ž ์•ฝ์นญ, 2. ์ œ๋ชฉ์˜ ๊ฒฝ์šฐ ๊ด€์‚ฌ์™€ ์ ‘์†์‚ฌ, ์ „์น˜์‚ฌ๋ฅผ ์ œ์™ธํ•œ ๋‹จ์–ด์™€ ์ œ๋ชฉ ์ฒซ ๋‹จ์–ด์˜ ์ฒซ๊ธ€์ž๋Š” ๋Œ€๋ฌธ์ž๋กœ ์ž‘์„ฑ)
  • list, dict ๋“ฑ ํŒŒ์ด์ฌ ๊ธฐ๋ณธ ์ž๋ฃŒํ˜•์˜ ๊ฒฝ์šฐ ๋ฐœ์Œ๋Œ€๋กœ ํ‘œ๊ธฐํ•˜๊ณ  ์›์–ด๋Š” ๋ณ‘๊ธฐํ•˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
  • int, float, integer ๋“ฑ ์ž๋ฃŒํ˜• ํ‚ค์›Œ๋“œ/๋‹จ์–ด์˜ ๊ฒฝ์šฐ
    • ๋ฌธ์žฅ ๋‚ด์— ๋“ฑ์žฅํ•˜๋Š” ๊ฒฝ์šฐ ํ•œ๊ตญ์–ด๋กœ ๋ฒˆ์—ญํ•ฉ๋‹ˆ๋‹ค. (์˜ˆ: "~ is tuple of integers" โ†’ "~๋Š” ์ •์ˆ˜ํ˜• ํŠœํ”Œ์ž…๋‹ˆ๋‹ค.")
    • argument๋“ฑ ๋ณ€์ˆ˜ ์„ค๋ช…์—์„œ ์ž…๋ ฅ๊ฐ’์˜ ์ž๋ฃŒํ˜•์„ ๋‚˜ํƒ€๋‚ด๋Š” ๊ฒฝ์šฐ highlight๋กœ ํ‘œ์‹œํ•˜๊ณ  ํŒŒ์ด์ฌ ์ž๋ฃŒํ˜• ํ‘œ๊ธฐ๋Œ€๋กœ ์ ์Šต๋‹ˆ๋‹ค. (์˜ˆ: X: Integer, โ†’ int.)
  • ๋ฌธ์žฅ ๋์˜ colon(:)์€ ๋งˆ์นจํ‘œ๋กœ ๋Œ€์ฒดํ•ฉ๋‹ˆ๋‹ค.
    • ๋ฌธ์žฅ ๋์˜ semicolon(;)์€ ๋ฌธ์žฅ์„ ๋‘ ๊ฐœ๋กœ ๋ถ„๋ฆฌํ•˜๊ณ  ํ•„์š”ํ•œ ๊ฒฝ์šฐ ์ ํ•ฉํ•œ ์ ‘์†์‚ฌ๋ฅผ ์ถ”๊ฐ€ํ•ฉ๋‹ˆ๋‹ค.
  • Keras๋ฅผ ์ œ์™ธํ•œ ๋ชจ๋“  API ๋ฐ ์„œ๋น„์Šค ๋“ฑ์˜ ์ด๋ฆ„(TensorFlow, NumPy, CNTK, Amazon, Google ๋“ฑ)์€ ์›๋ฌธ ๊ทธ๋Œ€๋กœ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค
  • ํ•จ์ˆ˜ ์ธ์ž ์„ค๋ช…์‹œ [์ธ์ž: data type, ์„ค๋ช… ๋‚ด์šฉ, ๊ธฐ๋ณธ๊ฐ’ ]์˜ ํ˜•์‹์„ ๋”ฐ๋ฆ…๋‹ˆ๋‹ค. (์˜ˆ: batch_size: int ํ˜น์€ None. ์†์‹ค๋กœ๋ถ€ํ„ฐ ๊ทธ๋ž˜๋””์–ธํŠธ๋ฅผ ๊ตฌํ•˜๊ณ  ๊ฐ€์ค‘์น˜๋ฅผ ์—…๋ฐ์ดํŠธํ•˜๋Š” ๊ณผ์ • ํ•œ ๋ฒˆ์— ์‚ฌ์šฉํ•  ํ‘œ๋ณธ์˜ ๊ฐœ์ˆ˜์ž…๋‹ˆ๋‹ค. ๊ธฐ๋ณธ๊ฐ’์€ 32์ž…๋‹ˆ๋‹ค.)
  • Raises๋ž€์˜ ๊ฒฝ์šฐ ์˜ค๋ฅ˜๋กœ ๋ฒˆ์—ญํ•˜๋ฉฐ, ๋ณธ๋ฌธ์€ "(~ํ•˜๋Š” ๊ฒฝ์šฐ, ~ํ•˜๋ฉด, ~๊ฐ€) ๋ฐœ์ƒํ•ฉ๋‹ˆ๋‹ค."๋กœ ์ •๋ฆฌํ•ฉ๋‹ˆ๋‹ค.

์šฉ์–ด ํ†ต์ผ์•ˆ

English ํ•œ๊ตญ์–ด
-er ~ํ™” ํ•จ์ˆ˜ / ํ•จ์ˆ˜
1--9 1--9
accuracy ์ •ํ™•๋„
argument ์ธ์ž
(artificial) neural network (์ธ๊ณต) ์‹ ๊ฒฝ๋ง
augmenter ์ฆ๊ฐ• ํ•จ์ˆ˜
Average Pooling ํ‰๊ท  ํ’€๋ง
axis ์ถ•
batch ๋ฐฐ์น˜
bias ํŽธํ–ฅ
binary classification ์ด์ง„ ๋ถ„๋ฅ˜
cache ์บ์‹œ
callback ์ฝœ๋ฐฑ
cell state ์…€ ์ƒํƒœ
channel ์ฑ„๋„
checkpoint ์ฒดํฌํฌ์ธํŠธ
class ํด๋ž˜์Šค
classification ๋ถ„๋ฅ˜
compile ์ปดํŒŒ์ผ
constraint ์ œ์•ฝ
convolutional neural network (CNN) ํ•ฉ์„ฑ๊ณฑ ์‹ ๊ฒฝ๋ง
corpus ๋ง๋ญ‰์น˜
dense layer ์™„์ „์—ฐ๊ฒฐ์ธต
dimension ์ฐจ์›
dot product ๋‚ด์ 
dropout ๋“œ๋กญ์•„์›ƒ
element-wise ์›์†Œ๋ณ„
embedding ์ž„๋ฒ ๋”ฉ
encoding ์ธ์ฝ”๋”ฉ
epoch ์—ํญ (์„œ์ˆ ์ ์œผ๋กœ ์“ธ ๋•Œ๋Š” 'nํšŒ ๋ฐ˜๋ณต')
factor ๊ฐ’/์š”์ธ/์š”์†Œ
fully-connected, densely connected ์™„์ „ ์—ฐ๊ฒฐ
global ์ „์—ญ
generator ์ œ๋„ˆ๋ ˆ์ดํ„ฐ
gradient ๊ทธ๋ž˜๋””์–ธํŠธ
gradient ascent ๊ฒฝ์‚ฌ์ƒ์Šน๋ฒ•
gradient descent ๊ฒฝ์‚ฌํ•˜๊ฐ•๋ฒ•
hidden unit ์€๋‹‰ ์œ ๋‹›
hidden layer ์€๋‹‰ ์ธต
hidden state ์€๋‹‰ ์ƒํƒœ
hyperparameter ํ•˜์ดํผํŒŒ๋ผ๋ฏธํ„ฐ
identity matrix ๋‹จ์œ„ ํ–‰๋ ฌ
index ์ธ๋ฑ์Šค (๊ฐœ๋ณ„ index์˜ ๋ฌถ์Œ ์ „์ฒด๋ฅผ ๊ฐ€๋ฆฌํ‚ฌ ๋•Œ๋Š” '๋ชฉ๋ก')
input ์ž…๋ ฅ/์ž…๋ ฅ๊ฐ’
instance ์ธ์Šคํ„ด์Šค
initialization ์ดˆ๊ธฐ๊ฐ’ ์ƒ์„ฑ
initializer ์ดˆ๊ธฐํ™” ํ•จ์ˆ˜
keras ์ผ€๋ผ์Šค
kernel ์ปค๋„
label ๋ ˆ์ด๋ธ”
layer ์ธต
learning rate ํ•™์Šต๋ฅ 
learning rate decay ํ•™์Šต๋ฅ  ๊ฐ์†Œ
locally ๋ถ€๋ถ„ ์—ฐ๊ฒฐ
loss function ์†์‹ค ํ•จ์ˆ˜
LSTM LSTM
MaxPooling ์ตœ๋Œ“๊ฐ’ ํ’€๋ง
mean squared error (MSE) ํ‰๊ท  ์ œ๊ณฑ ์˜ค์ฐจ(๋ฒ•)
metric (ํ‰๊ฐ€) ์ง€ํ‘œ (๋ฌธ๋งฅ์— ๋”ฐ๋ผ ์œ ์—ฐํ•˜๊ฒŒ ์‚ฌ์šฉ)
mini-batch ๋ฏธ๋‹ˆ ๋ฐฐ์น˜
model ๋ชจ๋ธ
momentum ๋ชจ๋ฉ˜ํ…€
multi-class classification ๋‹ค์ค‘ ๋ถ„๋ฅ˜
multilayer perceptron (MLP) ๋‹ค์ธต ํผ์…‰ํŠธ๋ก 
neuron ๋‰ด๋Ÿฐ
node ๋…ธ๋“œ
noise ๋…ธ์ด์ฆˆ
non-negativity ์Œ์ด ์•„๋‹Œ ~
norm ๋…ธ๋ฆ„
normalization ์ •๊ทœํ™”
normalize ์ •๊ทœํ™”ํ•˜๋‹ค
note ์ฐธ๊ณ 
objective function ๋ชฉ์  ํ•จ์ˆ˜
one-hot encoding ์›-ํ•ซ ์ธ์ฝ”๋”ฉ
optimizer ์ตœ์ ํ™” ํ•จ์ˆ˜
output ์ถœ๋ ฅ(๊ฐ’)
padding ํŒจ๋”ฉ
parameter (ํ•จ์ˆ˜์˜)๋งค๊ฐœ๋ณ€์ˆ˜
parameter (๋ชจ๋ธ์˜)ํŒŒ๋ผ๋ฏธํ„ฐ (๊ฐ€์ค‘์น˜์™€ ํŽธํ–ฅ์„ ํ•จ๊ป˜ ์ด๋ฅด๋Š” ๋ง)
placeholder ํ”Œ๋ ˆ์ด์Šคํ™€๋”
penalty ํŽ˜๋„ํ‹ฐ
pooling ํ’€๋ง
precision ์ •๋ฐ€๋„
queue ๋Œ€๊ธฐ์—ด
recurrent neural network (RNN) ์ˆœํ™˜ ์‹ ๊ฒฝ๋ง
reference ์ฐธ๊ณ 
regression ํšŒ๊ท€ ๋ถ„์„
regression(-ive) model ํšŒ๊ท€ ๋ชจ๋ธ
regularize(-er) ๊ทœ์ œํ™”/๊ทœ์ œ ํ•จ์ˆ˜
repository ์ €์žฅ์†Œ
reshape ํ˜•ํƒœ๋ฐ”๊พธ๊ธฐ
return ๋ฐ˜ํ™˜๊ฐ’
root mean squared error (RMSE) ํ‰๊ท  ์ œ๊ณฑ๊ทผ ์˜ค์ฐจ(๋ฒ•)
sample ํ‘œ๋ณธ
sequence (-tial) ์ˆœ์„œํ˜•
set ์„ธํŠธ
shape ํ˜•ํƒœ
stack ์ธต์„ ์Œ“๋‹ค
stateful ์ƒํƒœ ์ €์žฅ
stochastic gradient descent ํ™•๋ฅ ์  ๊ฒฝ์‚ฌํ•˜๊ฐ•๋ฒ•
stride ์ŠคํŠธ๋ผ์ด๋“œ
target ๋ชฉํ‘œ(๊ฐ’)
temporal ์‹œ๊ณ„์—ด
tensor ํ…์„œ
test ์‹œํ—˜
text ํ…์ŠคํŠธ
timestep ์‹œ๊ฐ„ ๋‹จ๊ณ„/์ˆœ์„œ
token ํ† ํฐ
train (๋ฐ์ดํ„ฐ์˜ ๊ฒฝ์šฐ) ํ›ˆ๋ จ ์„ธํŠธ / (๋™์ž‘์˜ ๊ฒฝ์šฐ) ํ•™์Šต์‹œํ‚ค๋‹ค
utility ๋„๊ตฌ
validation ๊ฒ€์ฆ
weight ๊ฐ€์ค‘์น˜
wrapper ๋ž˜ํผ