Basics of deep learning


This educational video is the third part of the complete course of deep learning in MATLAB.

In this educational video, we will introduce you to the basics of deep learning so that you can gain the necessary knowledge to work in the field of deep learning and be able to implement your models in different programming languages.

Artificial Intelligence machine learning deep learning MATLAB

Artificial intelligence is a simulation of human intelligence. Mankind’s long-held dream of being able to design a system that can function like human intelligence. The human brain, which is the core of decision-making, contains millions of neural neurons that are primarily responsible for intelligence. Every person acquires a huge amount of knowledge and data from their environment during their life. This data is processed by the brain and knowledge is extracted from them. Mathematically expressing the learning function of the brain from the environment is very difficult, so one of the difficulties in modeling intelligence in the brain is the lack of human understanding of the learning function of the brain and its transformation into intelligence. We need to design computer systems that capture the same input as the brain and turn it into intelligence. The key question here is how to turn external data into intelligence so that a computer system can train like our brain and function like our brain.

To solve this problem, the researchers introduced algorithms that extract patterns from the input data. This concept is known as machine learning. With the introduction of machine learning, intelligent systems were able to extract information from the external environment, and models were introduced that to some extent solve the previous problem in artificial intelligence. A simple machine learning algorithm is naïve Bayes that can separate spam emails from real emails.

The efficiency of machine learning algorithms depends on how the input data is expressed. This concept is known as representation. As a system, consider the diagnosis based on machine learning, which performs its diagnostic work based on the results of the patient’s tests such as the amount of sugar, urea, fat, etc. This system itself cannot communicate directly with the patient and can only perform diagnostic work based on the data that the doctor gives to the system. This means that if the doctor does not enter the number of blood lipids, the system cannot respond because it is one of the unknown inputs. This system can not work on the basis of the MRI image because it can not have a correct understanding of the image and only makes decisions based on the patient’s test data.

Many AI tasks can be solved by designing features to extract data from input with the help of machine learning algorithms. For example, one of the features used to identify the speaker in speech signal processing is estimating the size of the speaker’s vocal tract. An algorithm with this feature can detect whether the speaker is male, female, or child.

The important point here is that we do not know what features to extract in each user to achieve the best efficiency, so feature extraction is one of the weaknesses of machine learning models. Sometimes finding the right features for a particular application is found after realizing hundreds of researchers, which indicates the weakness of machine learning models.

Deep learning was proposed to solve the problem of machine learning models. Deep learning models no longer followed feature extraction algorithms and could function like the human brain with the help of their layers. You can see a conceptual form of deep learning in the following figure, which is well explained in the educational video of the basics of deep learning in MATLAB :

Basics of deep learning MATLAB

Basics of deep learning MATLAB


Headlines :

human brain

Neurons in the human brain

Why deep learning?

History of artificial intelligence and deep learning

The first artificial neural network

An overview of artificial intelligence and human competitions

Today’s applications of artificial intelligence

Example of a driverless car

Milestones in AI

Statistics of the number of articles in the field of deep learning

Introducing TensorFlow, Pytorch and Keras libraries

Relationship between Keras and TensorFlow

Introducing CPU, GPU and TPU

Deep learning programming for mobile

Deep learning programming for the cloud

Farewell to Python 2

The Importance of Python 3

What is machine learning?

How to implement machine learning?

The importance of data in deep learning

Why do we have to learn deep learning now?

Recent advances in deep learning

Deep learning in medicine

Diagnosis of genetic disease from a person’s face

Various applications of deep learning

The connection between artificial intelligence and machine learning and deep learning

The difference between machine learning and deep learning

The concept of deep learning

(Edge, corners, and parts of the object) In the visual machine with deep learning

A step-by-step explanation of a deep learning model

Interpret the output of a deep learning system

Why does a deep learning system output as a percentage of probability?

The concept of representation in deep learning

Increase data dimension

Relationship between efficiency and volume of training data in deep learning and machine learning

The role of data in deep learning

The disadvantage of deep learning

How much data is needed for a deep learning system?

A practical example from GitHub

Why is so much data needed in deep learning?

Despite its weakness in deep learning, why is it still so widely used?

Difficulties in deep learning

Inter class variation

Data argumentation

adversarial attack

A practical example of deep learning

Types of learning

Learn with the observer

Pseudo-learning with the observer

Reinforcement learning

Learning without an observer

Patterning of human learning

An example of amplified learning in driverless cars

Cut data

Data reflection

Resize data

Rotate the data

Data transfer

Noise in data

Challenges in deep learning

Training and testing

The concept of regression

The concept of classification

The concept of Multi-Class

Multi-Label Concept

What can we do with deep learning?

Deep concept

Hidden layer

Neural networks and deep learning

The concept of overfitting

Impact of data number on overfitting

How do we know if overfitting is happening?

Concept of Regularization

Validation data and training data

Use Dropout

Dropout layer


Batch normalization

Batch Renormalization

What is the purpose of normalization?

What is the meaning of batch?

What is the effect of batch normalization on deep learning?

Introducing famous deep learning networks

Introducing the AlexNet network

Introducing VGG-16 network

Introducing GoogleNet

Change predefined networks

An important point in using pre-trained deep learning networks

The evolution of deep learning networks

Can a deep learning model do the job of recognition better than humans?

Comparing the efficiency of Alexnet, GoogleNet, ZFNet, VGGNet, RESNet, GoogleNet v4 and SENet networks

object detection

Positioning objects

Semantic Segmentation

What is the heart of deep learning?

The concept of Transfer Learning

Generative Adversarial Network (GAN)

Advantage of GAN network

GAN network structure

Generator and Discriminator function

Introduction to AutoML machine learning

Deep learning network design automatically

NASNet Network

Search for the best structure for deep learning

Introducing Reinforcement Learning

The concept of Agent and Environment

Depth measurement in deep learning models

RNN return networks

Advantages and disadvantages of RNN network





There are no reviews yet.

Be the first to review “Basics of deep learning”

Your email address will not be published. Required fields are marked *