[PYTHON] Deep Learning from scratch Chapter 2 Perceptron (reading memo)

Introduction

What is this article?

It is a material and memo writing for the study session. We cannot guarantee the validity of the content, so we recommend that you refrain from quoting.

Reference link

-[Perceptron-Wikipedia](https://ja.wikipedia.org/wiki/%E3%83%91%E3%83%BC%E3%82%BB%E3%83%97%E3%83%88% E3% 83% AD% E3% 83% B3) -Machine learning that even high school graduates can understand (2) Simple perceptron | When you think of it in your head -Machine learning that even high school graduates can understand (3) Multilayer perceptron | When you think of it in your head -15th Classification Problem Beginning: Let's Start Machine Learning | gihyo.jp… Technical Review Company -Machine learning learned by Yaruo-Perceptron --- Kengo's mansion

2.1 What is Perceptron?

perceptron.png

_2 Input perceptron (from "Deep Learning from scratch" P.22) _

--A model that receives multiple signals as inputs and outputs one signal --The signal is a binary value of 0 (do not flow) or 1 (flow) --Each input is multiplied by a unique weight $ w ^ 1, w ^ 2 ... $ (e.g. $ w ^ 1x ^ 1 $) ――The higher the weight, the more important the target signal becomes. --Outputs 1 when the sum of the input values exceeds an arbitrary threshold $ \ theta $ ("neurons fire")

Assuming that the output is $ y $, the above elements can be expressed mathematically as follows.

y = \left\{
\begin{array}{ll}
0 & (w^1x^1 + w^2x^2 \leqq \theta) \\
1 & (w^1x^1 + w^2x^2 \gt \theta)
\end{array}
\right.

2.2 Simple logic circuit

--The AND / NAND / OR logic circuit can be expressed by using the perceptron. --Change the behavior of changing weights and thresholds without changing the structure of the perceptron

2.2.1 AND gate

_P.23 See "Fig. 2-2 AND Gate Truth Table" _

--Outputs 1 only when two inputs are 1, otherwise outputs 0 --Perceptron that outputs when the sum of two input values exceeds the threshold $ \ theta $

2.2.2.1 NAND gate

_P.24 See "Figure 2-3 NAND Gate Truth Table" _

--NAND = Not AND = AND Behavior opposite to gate --Outputs 0 only when both are 1, and outputs 1 otherwise --When the sign of the parameter that realizes the AND gate is inverted, it becomes a NAND gate.

2.2.2.2 OR gate

--Output 1 if at least one input is 1.

2.3 Implementation of Perceptron

2.3.1 Easy implementation

Let's define and execute the AND circuit as follows.

def AND(x1, x2):
  w1, w2, theta = 0.5, 0.5, 0.7
  tmp = x1 * w1 + x2 * w2

  if tmp <= theta:
    return 0
  elif tmp > theta:
    return 1

2.3.2 Introducing weights and biases

--Introduce a bias into the AND circuit of 2.3.1. --Replace $ \ theta $ in Equation 2.1 with $ -b $, where the bias is $ b $.

y = \left\{
\begin{array}{ll}
0 & (w^1x^1 + w^2x^2 \leqq \theta) \\
1 & (w^1x^1 + w^2x^2 \gt \theta)
\end{array}
\right.
y = \left\{
\begin{array}{ll}
0 & (b + w^1x^1 + w^2x^2 \leqq 0) \\
1 & (b + w^1x^1 + w^2x^2 \gt 0)
\end{array}
\right.

Let's check with the interpreter.

>>> import numpy as np
>>> x = np.array([0, 1]) #input
>>> w = np.array([0.5, 0.5]) #weight
>>> b = -0.7 #bias
>>> w * x
>>> np.sum(w * x)
>>> np.sum(w * x) + b

2.3.3 Implementation by weight and bias

Let's implement each circuit of AND / NAND / OR based on the previous section.

def AND(x1, x2):
  x = np.array([x1, x2])

  w = np.array([0.5, 0.5])
  b = -0.7

  tmp = np.sum(w * x) + b
  
  if tmp <= 0:
    return 0
  else:
    return 1
def NAND(x1, x2):
  x = np.array([x1, x2])

  #Only weights and biases differ from AND
  w = np.array([-0.5, -0.5])
  b = 0.7

  tmp = np.sum(w * x) + b
  
  if tmp <= 0:
    return 0
  else:
    return 1
def OR(x1, x2):
  x = np.array([x1, x2])

  #Only bias is different from AND
  w = np.array([0.5, 0.5])
  b = -0.2

  tmp = np.sum(w * x) + b
  
  if tmp <= 0:
    return 0
  else:
    return 1

Difference between bias and weight

--Weights $ w ^ 1, w ^ 2 ... $ control the importance of the input signal --Bias adjusts the ease of firing of neurons (probability of outputting 1)

The term bias means "geta haki". This means how much geta (adding a value) to the output when there is no input (when the input is 0). In fact, the calculation of b + w1 x1 + w2 x2 in Equation (2.2) prints only the bias value if the inputs x1 and x2 are 0.

(Excerpt from P.27)

2.4 Limits of Perceptron

--A single-layer perceptron can separate linear regions, but not non-linear regions such as XOR (exclusive OR).

2.5 Multilayer Perceptron

--XOR can be expressed by "layering". --A perceptron with multiple layers is called a ** multi-layered perceptron **, and is expressed as "$ n $ layer perceptron".

2.5.1 Combination of existing gates

An XOR circuit that outputs 1 when one of the inputs is 1 and the other is 0 is represented by a combination of existing AND / NAND / OR. Let's sort out the behavior of each circuit again.

--AND ... Outputs 1 if the two inputs are 1, otherwise outputs 0. --NAND ... Outputs 0 if the two inputs are 1, otherwise outputs 1. --Outputs 1 if at least one of the two inputs is 1, otherwise outputs 0. --XOR ... Outputs 1 if one of the two inputs is 1 and the other is 0, otherwise outputs 0.

Actually, it can be realized by the following wiring. The initial value is $ x ^ 1, x ^ 2 $, the NAND output is $ s ^ 1 $, and the OR output is $ s ^ 2 $.

xor.png _P.32 Figure 2-11_

Check the truth table.

xor-chart.png _P.33 Figure 2-12 XOR Gate Truth Table _

2.5.2 XOR gate implementation

Let's implement it in the code based on the previous section.

def XOR(x1, x2):
  s1 = NAND(x1, x2)
  s2 = OR(x1, x2)
  y = AND(s1, s2)
  return y

2.6 NAND to computer

2.7 Summary

--Perceptron is an algorithm with input and output. Given a certain input, a fixed value is output. --In Perceptron, "weight" and "bias" are set as parameters. --By using Perceptron, you can express logic circuits such as AND and OR gates. Wear. --The XOR gate cannot be represented by a single-layer perceptron. --The XOR gate can be represented by using a two-layer perceptron. --Single-layer perceptrons can only represent linear regions, whereas multi-layer perceptrons Perceptron can represent a non-linear region. --A multi-layered perceptron can (in theory) represent a computer.

Recommended Posts

Deep Learning from scratch Chapter 2 Perceptron (reading memo)
[Learning memo] Deep Learning made from scratch [Chapter 7]
Deep learning / Deep learning made from scratch Chapter 6 Memo
[Learning memo] Deep Learning made from scratch [Chapter 5]
[Learning memo] Deep Learning made from scratch [Chapter 6]
Deep learning / Deep learning made from scratch Chapter 7 Memo
[Learning memo] Deep Learning made from scratch [~ Chapter 4]
Deep Learning / Deep Learning from Zero 2 Chapter 4 Memo
Deep Learning / Deep Learning from Zero Chapter 3 Memo
Deep Learning / Deep Learning from Zero 2 Chapter 5 Memo
Deep Learning / Deep Learning from Zero 2 Chapter 7 Memo
Deep Learning / Deep Learning from Zero 2 Chapter 8 Memo
Deep Learning / Deep Learning from Zero Chapter 5 Memo
Deep Learning / Deep Learning from Zero Chapter 4 Memo
Deep Learning / Deep Learning from Zero 2 Chapter 3 Memo
Deep Learning / Deep Learning from Zero 2 Chapter 6 Memo
Deep Learning from scratch
Learning record of reading "Deep Learning from scratch"
"Deep Learning from scratch" Self-study memo (Part 12) Deep learning
"Deep Learning from scratch" self-study memo (unreadable glossary)
Deep Learning from scratch ① Chapter 6 "Techniques related to learning"
[Learning memo] Deep Learning from scratch ~ Implementation of Dropout ~
"Deep Learning from scratch" Self-study memo (10) MultiLayerNet class
"Deep Learning from scratch" Self-study memo (No. 11) CNN
Deep Learning from scratch 1-3 chapters
Python vs Ruby "Deep Learning from scratch" Chapter 2 Logic circuit by Perceptron
"Deep Learning from scratch 2" Self-study memo (No. 21) Chapters 3 and 4
Deep learning from scratch (cost calculation)
Deep Learning memos made from scratch
An amateur stumbled in Deep Learning from scratch Note: Chapter 1
Making from scratch Deep Learning ❷ An amateur stumbled Note: Chapter 5
Making from scratch Deep Learning ❷ An amateur stumbled Note: Chapter 2
An amateur stumbled in Deep Learning from scratch Note: Chapter 3
An amateur stumbled in Deep Learning from scratch Note: Chapter 7
An amateur stumbled in Deep Learning from scratch Note: Chapter 5
Making from scratch Deep Learning ❷ An amateur stumbled Note: Chapter 7
Making from scratch Deep Learning ❷ An amateur stumbled Note: Chapter 1
Making from scratch Deep Learning ❷ An amateur stumbled Note: Chapter 4
"Deep Learning from scratch" self-study memo (No. 18) One! Meow! Grad-CAM!
"Deep Learning from scratch" self-study memo (No. 19-2) Data Augmentation continued
An amateur stumbled in Deep Learning from scratch Note: Chapter 4
An amateur stumbled in Deep Learning from scratch Note: Chapter 2
I tried to implement Perceptron Part 1 [Deep Learning from scratch]
"Deep Learning from scratch" self-study memo (No. 15) TensorFlow beginner tutorial
Making from scratch Deep Learning ❷ An amateur stumbled Note: Chapter 6
Deep learning from scratch (forward propagation edition)
Deep learning / Deep learning from scratch 2-Try moving GRU
"Deep Learning from scratch" in Haskell (unfinished)
[Windows 10] "Deep Learning from scratch" environment construction
[Deep Learning from scratch] About hyperparameter optimization
"Deep Learning from scratch" self-study memo (No. 13) Try using Google Colaboratory
"Deep Learning from scratch" Self-study memo (No. 10-2) Initial value of weight
"Deep Learning from scratch" Self-study memo (No. 14) Run the program in Chapter 4 on Google Colaboratory
"Deep Learning from scratch" Self-study memo (Part 8) I drew the graph in Chapter 6 with matplotlib
Chapter 2 Implementation of Perceptron Cut out only the good points of deep learning made from scratch
Good book "Deep Learning from scratch" on GitHub
Django memo # 1 from scratch
Python vs Ruby "Deep Learning from scratch" Summary
Python vs Ruby "Deep Learning from scratch" Chapter 4 Implementation of loss function
Python vs Ruby "Deep Learning from scratch" Chapter 3 Implementation of 3-layer neural network
Deep Learning from scratch The theory and implementation of deep learning learned with Python Chapter 3