Deep Learning - Recurrent Neural Networks in Python

發表於 更新於 分類於 Disqus：
文章字數： 4.2k 所需閱讀時間 ≈ 4 分鐘

說明

名詞

Recurrent neural network(循環神經網路 RNNs) :
Linear and Logistic Regression(線性邏輯回歸)
Backpropagation(反向傳播)
TensorFlow 2
Time series forecasting(時間序列預測)
Text classification(link sentiment analysis) 文字分類(情緒分析)
image recognition (圖像辨識)
LSTMs

Basic

Classification 分類
Regression 回歸
Neurons 神經元
Scikit-learn : 是用於Python程式語言的自由軟體機器學習庫

Numpy

Matrix arithmetic
Tensor(arrays)
1-D tensor(vector), 2-D tensoer(matrix)
Arithmetic:+,-,,/
Matrix mutiply == dot/inner product(np.dot)
Element-wise multiply()

Matplotlib

Line charts
Scatterplots

Pandas

Loading in tabular data

Scipy

Is like a power version od Numpy(statistics, optimization, alinear algebra, signal processing)
Numpy is lower-level:adding, multiplying

Scikit-Learn

It’s about the concept behind using it
Classicfication and regression
MAchine learning as geometry rather than magic
Shape of data(X and Y)

The Step of a Machine Learning Script

  • Load in the data
    • X and Y
      • Typically use PAndas usless the data it too complex
  • Slipt train/test sets(sometimes)
    • Sometimes “test” and “validation” are used interchangeably, and the “true test set” is sometiming else - don’t get frazzled over this
  • Build a Model
    • OOP(object oriented programing)
    • TensorFlow 2.0 standard is Keras API, very similar to Sciki-Learn
  • Fit the model(gradient descent)
    • model.fit(X,Y)
  • Evaluate the model
  • Make predictions
    • model.predict(X)

Google Colab

setting

CPU, GPU

執行階段 –> 變更執行階段類型

uploae/download notebook file

檔案 –> 上傳筆記本
檔案 –> 下載

command
1
2
3
4
5
# runs shell
!ls
!pip list
#! sed ..
#! unzip file

magic command

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
# show magic command
%lsmagic

# change director
# !cd not work
%cd dd

# 計算單行執行時間
%timeit a=1+2+3
# 20.1 ns ± 3.89 ns per loop (mean ± std. dev. of 7 runs, 100000000 loops each)

# 計算多行執行時間
%%timeit
sum=0
for i in range(1, 10000):
  sum += i
# 803 µs ± 202 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)	

# run one file
#%fun file 

# see variable 
a = b = c = 100
%whos 
# Variable   Type    Data/Info
# ----------------------------
# a          int     100
# b          int     100
# c          int     100


Upload

Using wget
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
# download the data from a URL
!wget https://lazyprogrammer.me/course_files/arrhythmia.data

# list files in current directory
!ls

# check if the data has a header
!head arrhythmia.data

# check the data
import pandas as pd
df = pd.read_csv('arrhythmia.data', header=None)
# since the data has many columns, take just the first few and name them (as per the documentation)
data = df[[0,1,2,3,4,5]]
data.columns = ['age', 'sex', 'height', 'weight', 'QRS duration', 'P-R interval']

import matplotlib.pyplot as plt
plt.rcParams['figure.figsize'] = [15, 15] # make the plot bigger so the subplots don't overlap
data.hist(); # use a semicolon to supress return value

from pandas.plotting import scatter_matrix
scatter_matrix(data);
Using tf.keras
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# use keras get_file to download the auto MPG dataset
### alternate URL
url = 'https://raw.githubusercontent.com/lazyprogrammer/machine_learning_examples/master/tf2.0/auto-mpg.data'

import tensorflow as tf
print(tf.__version__)

# check out the documentation for other arguments
tf.keras.utils.get_file('auto-mpg.data', url)

!head /root/.keras/datasets/auto-mpg.data

# unless you specify an alternative path, the data will go into /root/.keras/datasets/
df = pd.read_csv('/root/.keras/datasets/auto-mpg.data', header=None, delim_whitespace=True)
df.head()
Upload the file yourself
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
# another method: upload your own file
# if you must, then get the file from here:
# https://raw.githubusercontent.com/lazyprogrammer/machine_learning_examples/master/tf2.0/daily-minimum-temperatures-in-me.csv
from google.colab import files
uploaded = files.upload()

uploaded

# file is uploaded to the current directory
!ls

# open the file
# the last few lines are junk
df = pd.read_csv('daily-minimum-temperatures-in-me.csv', error_bad_lines=False)
df.head()

# upload a Python file with some useful functions (meant for fake_util.py)
from google.colab import files
uploaded = files.upload()

from fake_util import my_useful_function
my_useful_function()

!pwd
1
2
3
# fake_util.py
def my_useful_function():
  print("hello world")
Upload the file yourself
1
2
3
4
5
6
7
8
9
10
11
12
# Access files from your Google Drive
from google.colab import drive
drive.mount('/content/gdrive')

# Check current directory - now gdrive is there
!ls

# What's in gdrive?
!ls gdrive

# Whoa! Look at all this great VIP content!
!ls /content/gdrive/MyDrive

Ref