Note_Tech

All technological notes.

---

Project maintained by simonangel-fong Hosted on GitHub Pages — Theme by mattgraham

PyTorch - Perceptron

Back

• PyTorch - Perceptron
  • Perceptron
  • Example(待续)

---

Perceptron

• Perceptron / Linear Binary Classifier.

  • a single layer neural network
  • a binary classifier used in supervised learning

• multi-layer perceptron = a neural network

• binary classifiers

  • A function that can decide whether or not an input which is represented by a vector of number belongs to some specific class. 是否某个类.

---

• Perceptron consist of four parts:
  • Input values / one input layer
  • Weights and bias
    • Weight
      • represents the strength or dimension of the connection between units.
      • determined the influence of the input will have on the output
      • shows the strength of the particular node.
    • Bias
      • similar to the intercept added in a linear equation.
      • an additional parameter to adjust the output along with the weighted sum of the inputs to the neuron.
      • A bias value allows you to shift the activation function curve up or down. 平移
  • Activation Function
    • determined whether a neuron is activated or not
    • calculates a weighted sum and further adding bias with it to give the result.
    • used to map the input between the required value like (0, 1) or (-1, 1)

---

• steps:
  • 1. all the input x are multiplied with their weights, denoted as K.
  • 1. calculate weighted sum:
       • add all the multiplied value from K1 to Kn
  • 1. applied to the correct activation function.

---

• Since perception returns a binary output,
  • loss function = cross-entropy()
  • optimizer = sigmoid gradient descent

---

Example(待续)

• creating the dataset

import torch
import numpy as np
import matplotlib.pyplot as plt
from sklearn import datasets

# define number of points
no_of_points = 100
# define clusters' centers
centers = [[-0.5, 0.5], [0.5, -0.5]]
# .make_blobs(): Generate isotropic Gaussian blobs for clustering.
x, y = datasets.make_blobs(
    n_samples=no_of_points,
    random_state=42,
    centers=centers,
    cluster_std=0.4
)

print(x)
print(y)

• Plot data

def Scatter():
    plt.scatter(x[y == 0, 0], x[y == 0, 1])
    plt.scatter(x[y == 1, 0], x[y == 1, 1])

Scatter()

---

TOP