SlideShare a Scribd company logo
2
Most read
4
Most read
Sigmoid Function: Machine
Learning Made Simple
Devansh
Overview
Actually a family of functions. All
functions have a characteristic "S"-
shaped curve or sigmoid curve.
The most famous example is the
logistic function. Other big ones are
tanh and arc tan.
By their nature, they can be used to
“squish” input, making them useful
in Machine Learning
Why and When we use the Sigmoid
As mentioned earlier, sigmoids can “squish” input. They can take input with very
high values and map it to a much more manageable range.
In neural networks, we see them being used as activation functions.
We can also put them in the last layer the network, converting model output into
probabilities. This is done to make the output easier to understand and interpret.
Sigmoid functions are often taught hand in hand with logistic regression. The
logistic function is a sigmoid.
Example of some sigmoid functions
Sigmoid and the Vanishing Gradient(why we have a ReLU)
Since the value of a sigmoid function converges to a limit, we know that the
derivative of the function converges to 0.
This can lead to the vanishing gradient problem (remember what it is?).
Also, 0 as a derivative makes training very slow. How can we work around this?
In modern artificial neural networks, it is common to see in place of the sigmoid
function, the rectifier, also known as the rectified linear unit, or ReLU, being used
as the activation function. The ReLU is defined as:
f(x)= max(0,x)
Some more about ReLU
Supposed to be closer to biological
reality (doesn’t get triggered for values
below a threshold).
Simple function->Fast Calculation.
Since the gradient is always 1, we avoid
both the vanishing gradient and the
learning rate slowdown.
Depending on the context, we can add a
slight bias to the gradient for our needs.
Reach out to me
Check out my other articles on Medium. : https://machine-learning-made-
simple.medium.com/
My YouTube: https://rb.gy/88iwdd
Reach out to me on LinkedIn: https://www.linkedin.com/in/devansh-devansh-516004168/
My Instagram: https://rb.gy/gmvuy9
My Twitter: https://twitter.com/Machine01776819
My Substack: https://devanshacc.substack.com/
Live conversations at twitch here: https://rb.gy/zlhk9y
Get a free stock on Robinhood: https://join.robinhood.com/fnud75

More Related Content

PPTX
Cloud Computing Security
PPTX
Challenges in Cloud Forensics
PPTX
Network attacks
PPTX
Cybersecurity Attack Vectors: How to Protect Your Organization
PPTX
Remote server
PDF
Password Cracking using dictionary attacks
PPTX
Firewall presentation
PPSX
Cloud computing information management
Cloud Computing Security
Challenges in Cloud Forensics
Network attacks
Cybersecurity Attack Vectors: How to Protect Your Organization
Remote server
Password Cracking using dictionary attacks
Firewall presentation
Cloud computing information management

What's hot (20)

PPTX
CLOUD STORAGE.pptx
PDF
Training Neural Networks
PPTX
DoS or DDoS attack
PPT
Client-Server Computing
PPTX
Detection of Phishing Websites
PPT
CYBER CRIME AND SECURITY
PDF
Process injection - Malware style
PPTX
Database security
PDF
Software Engineering - chp8- deployment
PPTX
Traditioanal vs-cloud based Data Centers
PPTX
Security of the database
PPTX
DDoS - Distributed Denial of Service
PPTX
cloud storage ppt
PDF
DeepLocker - Concealing Targeted Attacks with AI Locksmithing
PPTX
Introduction to Cyber Forensics Module 1
PPSX
Introduction to Cloud computing
PDF
Visual Question Answering 2.0
PDF
SDN-ppt-new
PPTX
Unit-I Introduction to Cloud Computing.pptx
PPTX
Traditional Problems Associated with Computer Crime
CLOUD STORAGE.pptx
Training Neural Networks
DoS or DDoS attack
Client-Server Computing
Detection of Phishing Websites
CYBER CRIME AND SECURITY
Process injection - Malware style
Database security
Software Engineering - chp8- deployment
Traditioanal vs-cloud based Data Centers
Security of the database
DDoS - Distributed Denial of Service
cloud storage ppt
DeepLocker - Concealing Targeted Attacks with AI Locksmithing
Introduction to Cyber Forensics Module 1
Introduction to Cloud computing
Visual Question Answering 2.0
SDN-ppt-new
Unit-I Introduction to Cloud Computing.pptx
Traditional Problems Associated with Computer Crime
Ad

Similar to Sigmoid function machine learning made simple (20)

PDF
NNAF_DRK.pdf
PPTX
Lecture02_Updated_Shallow Neural Networks.pptx
PDF
What are activation functions and why do we need those.pdf
PDF
APznzaY3lgBI1_4Em3SL-evNrENZMqxOBcBfS-4LCLWoo0zOcyDIrRDXBMqCS3EPzJW34i1165zNf...
PPTX
Activation functions
PPTX
activation functions activation functions.pptx
PPTX
Neural Networks by Priyanka Kasture
PPTX
Activation function
PDF
EuroAD 2021: ChainRules.jl
PPTX
OOP-Advanced Programming with c++
PDF
Understanding Deep Learning & Parameter Tuning with MXnet, H2o Package in R
PPTX
V2.0 open power ai virtual university deep learning and ai introduction
PDF
Programming Laboratory Unit 1.pdf
PDF
Machine Learning
PPTX
PRML Chapter 5
PDF
DNN.pdf (neural network basics and explanation|)
PDF
Folding Unfolded - Polyglot FP for Fun and Profit - Haskell and Scala - Part 2
PDF
Folding Unfolded - Polyglot FP for Fun and Profit - Haskell and Scala Part 2 ...
PPTX
4. OPTIMIZATION NN AND FL.pptx
PPTX
[7] The SiLU Activation Function Unlocking Neural Network Potential.pptx
NNAF_DRK.pdf
Lecture02_Updated_Shallow Neural Networks.pptx
What are activation functions and why do we need those.pdf
APznzaY3lgBI1_4Em3SL-evNrENZMqxOBcBfS-4LCLWoo0zOcyDIrRDXBMqCS3EPzJW34i1165zNf...
Activation functions
activation functions activation functions.pptx
Neural Networks by Priyanka Kasture
Activation function
EuroAD 2021: ChainRules.jl
OOP-Advanced Programming with c++
Understanding Deep Learning & Parameter Tuning with MXnet, H2o Package in R
V2.0 open power ai virtual university deep learning and ai introduction
Programming Laboratory Unit 1.pdf
Machine Learning
PRML Chapter 5
DNN.pdf (neural network basics and explanation|)
Folding Unfolded - Polyglot FP for Fun and Profit - Haskell and Scala - Part 2
Folding Unfolded - Polyglot FP for Fun and Profit - Haskell and Scala Part 2 ...
4. OPTIMIZATION NN AND FL.pptx
[7] The SiLU Activation Function Unlocking Neural Network Potential.pptx
Ad

More from Devansh16 (14)

PDF
Spine net learning scale permuted backbone for recognition and localization
PDF
Accounting for variance in machine learning benchmarks
PDF
When deep learners change their mind learning dynamics for active learning
PPTX
Semi supervised learning machine learning made simple
PDF
Paper Annotated: SinGAN-Seg: Synthetic Training Data Generation for Medical I...
PDF
A simple framework for contrastive learning of visual representations
PPTX
Recurrence relationships
PDF
Paper Explained: Deep learning framework for measuring the digital strategy o...
PDF
Paper Explained: One Pixel Attack for Fooling Deep Neural Networks
PDF
Paper Explained: Understanding the wiring evolution in differentiable neural ...
PPTX
Machine Learning Made Simple: Differential evolution
PDF
Paper Explained: RandAugment: Practical automated data augmentation with a re...
PDF
Noisy student images
PDF
Deep learning ensembles loss landscape
Spine net learning scale permuted backbone for recognition and localization
Accounting for variance in machine learning benchmarks
When deep learners change their mind learning dynamics for active learning
Semi supervised learning machine learning made simple
Paper Annotated: SinGAN-Seg: Synthetic Training Data Generation for Medical I...
A simple framework for contrastive learning of visual representations
Recurrence relationships
Paper Explained: Deep learning framework for measuring the digital strategy o...
Paper Explained: One Pixel Attack for Fooling Deep Neural Networks
Paper Explained: Understanding the wiring evolution in differentiable neural ...
Machine Learning Made Simple: Differential evolution
Paper Explained: RandAugment: Practical automated data augmentation with a re...
Noisy student images
Deep learning ensembles loss landscape

Recently uploaded (20)

PPTX
mbdjdhjjodule 5-1 rhfhhfjtjjhafbrhfnfbbfnb
PDF
Data Engineering Interview Questions & Answers Cloud Data Stacks (AWS, Azure,...
PPTX
Leprosy and NLEP programme community medicine
PPTX
Managing Community Partner Relationships
PPTX
STERILIZATION AND DISINFECTION-1.ppthhhbx
PDF
Transcultural that can help you someday.
PDF
Capcut Pro Crack For PC Latest Version {Fully Unlocked 2025}
PPTX
01_intro xxxxxxxxxxfffffffffffaaaaaaaaaaafg
PPTX
IB Computer Science - Internal Assessment.pptx
PPTX
Supervised vs unsupervised machine learning algorithms
PDF
annual-report-2024-2025 original latest.
PDF
Introduction to Data Science and Data Analysis
PDF
Introduction to the R Programming Language
PPTX
Microsoft-Fabric-Unifying-Analytics-for-the-Modern-Enterprise Solution.pptx
PDF
BF and FI - Blockchain, fintech and Financial Innovation Lesson 2.pdf
PPTX
Introduction to Firewall Analytics - Interfirewall and Transfirewall.pptx
PPTX
Introduction to Knowledge Engineering Part 1
PPT
Miokarditis (Inflamasi pada Otot Jantung)
PPTX
STUDY DESIGN details- Lt Col Maksud (21).pptx
PPTX
IBA_Chapter_11_Slides_Final_Accessible.pptx
mbdjdhjjodule 5-1 rhfhhfjtjjhafbrhfnfbbfnb
Data Engineering Interview Questions & Answers Cloud Data Stacks (AWS, Azure,...
Leprosy and NLEP programme community medicine
Managing Community Partner Relationships
STERILIZATION AND DISINFECTION-1.ppthhhbx
Transcultural that can help you someday.
Capcut Pro Crack For PC Latest Version {Fully Unlocked 2025}
01_intro xxxxxxxxxxfffffffffffaaaaaaaaaaafg
IB Computer Science - Internal Assessment.pptx
Supervised vs unsupervised machine learning algorithms
annual-report-2024-2025 original latest.
Introduction to Data Science and Data Analysis
Introduction to the R Programming Language
Microsoft-Fabric-Unifying-Analytics-for-the-Modern-Enterprise Solution.pptx
BF and FI - Blockchain, fintech and Financial Innovation Lesson 2.pdf
Introduction to Firewall Analytics - Interfirewall and Transfirewall.pptx
Introduction to Knowledge Engineering Part 1
Miokarditis (Inflamasi pada Otot Jantung)
STUDY DESIGN details- Lt Col Maksud (21).pptx
IBA_Chapter_11_Slides_Final_Accessible.pptx

Sigmoid function machine learning made simple

  • 2. Overview Actually a family of functions. All functions have a characteristic "S"- shaped curve or sigmoid curve. The most famous example is the logistic function. Other big ones are tanh and arc tan. By their nature, they can be used to “squish” input, making them useful in Machine Learning
  • 3. Why and When we use the Sigmoid As mentioned earlier, sigmoids can “squish” input. They can take input with very high values and map it to a much more manageable range. In neural networks, we see them being used as activation functions. We can also put them in the last layer the network, converting model output into probabilities. This is done to make the output easier to understand and interpret. Sigmoid functions are often taught hand in hand with logistic regression. The logistic function is a sigmoid.
  • 4. Example of some sigmoid functions
  • 5. Sigmoid and the Vanishing Gradient(why we have a ReLU) Since the value of a sigmoid function converges to a limit, we know that the derivative of the function converges to 0. This can lead to the vanishing gradient problem (remember what it is?). Also, 0 as a derivative makes training very slow. How can we work around this? In modern artificial neural networks, it is common to see in place of the sigmoid function, the rectifier, also known as the rectified linear unit, or ReLU, being used as the activation function. The ReLU is defined as: f(x)= max(0,x)
  • 6. Some more about ReLU Supposed to be closer to biological reality (doesn’t get triggered for values below a threshold). Simple function->Fast Calculation. Since the gradient is always 1, we avoid both the vanishing gradient and the learning rate slowdown. Depending on the context, we can add a slight bias to the gradient for our needs.
  • 7. Reach out to me Check out my other articles on Medium. : https://machine-learning-made- simple.medium.com/ My YouTube: https://rb.gy/88iwdd Reach out to me on LinkedIn: https://www.linkedin.com/in/devansh-devansh-516004168/ My Instagram: https://rb.gy/gmvuy9 My Twitter: https://twitter.com/Machine01776819 My Substack: https://devanshacc.substack.com/ Live conversations at twitch here: https://rb.gy/zlhk9y Get a free stock on Robinhood: https://join.robinhood.com/fnud75