Deep Learning Neural Network (CNN/RNN/GAN) Algorithm Principle + Practice

Ultra clear original painting complete without dense MP4 format supports any device

Download address:Baidu cloud disk

In this course, convolutional neural network (CNN), recurrent neural network (RNN) and adversarial neural network (GAN) in deep learning are explained in a simple way by means of principle explanation and actual practice. Through image classification, text classification, image style conversion, image text generation, image translation and other projects, students can acquire the ability to flexibly use CNN, RNN, GAN, the ability to tune deep learning algorithm and the ability to use Tensorflow for programming, and improve the ability of deep learning algorithm and project development experience.

** ** If you already know a programming language and want to become an ai engineer ** **, or if you are an undergraduate who wants to learn deep learning systematically at ** **, or if you are a beginner who lacks practical experience, then this course is perfect for you **

Fundamentals of Python fundamentals of probability and linear algebra Fundamentals of TensorFlow Fundamentals of machine learning

** Chapter contents: **
**

Chapter 1 Course Introduction
The guidance course of deep learning mainly introduces the application scope, talent demand and main algorithm of deep learning. It introduces the course chapters, course arrangement, applicable groups, prerequisite conditions and the degree achieved after learning, so that students can have a basic understanding of this course.
1-1 Course guide and Trial
Chapter 2 introduction to neural networks
The introductory course of this practical course. The introduction of machine learning and deep learning was made, and the latest progress of deep learning was explained through several examples. By explaining and practicing the basic structure of neural network — neuron and its extended Logistic regression model, the basic knowledge of this course is comprehensively explained, including neuron, activation function, objective function, gradient descent, learning rate, Tensorflow foundation and Tensorflow code implementation of the model. .
2-1 Introduction to machine learning and deep learning
2-2 neuron-Logistic regression model
2-3 neurons with multiple outputs
2-4 gradient descent
2-5 Data Processing and Model Graph Construction (1)
2-6 Data Processing and Model Graph Construction (2)
2-7 Neuronal realization (Realization of binary logistic regression model)
2-8 Realization of neural network (Realization of Multi-classification Logistic regression Model)
Chapter 3 Convolutional neural network
This course consists of two parts. The first part gives a complete introduction to neural network, including neural network structure, forward propagation, back propagation, gradient descent, etc. The second part explains the basic structure of convolutional neural network, including convolution, pooling and full connection. This paper focuses on the details of convolution operation, including convolution kernel structure, convolution calculation, and the number of convolution kernel parameters calculation, and introduces a basic convolution neural network structure. .
3-1 Advanced neural network
3-2 Convolutional neural network (1
Convolutional Neural Network (2)
3-4 Practical convolutional neural network
Chapter 4 advances of convolutional neural networks
This course introduces advanced convolutional neural network structures, including AlexNet, VGGNet, ResNet, InceptionNet, MobileNet, and their evolution process. For each structure, the course explains the problem it solves, the basic idea of the substructure, and the important techniques used in the model. After learning this course, students can achieve the ability to flexibly build different types of convolutional neural networks. .
4-1 Advanced Convolutional Neural Network (AlexNET)
4-2 Advanced Convolutional Neural Network (VGGNET-RESNET)
4-3 Advanced Convolutional Neural Network (Inception -mobile- Net)
Vgg-resnet (1)
Vgg-resnet (2)
Inception- Mobile_net (1)
Inception- Mobile_net (2)
Chapter 5 parameter tuning of convolutional neural network
This class systematically summarizes and summarizes the commonly used parameter tuning techniques (” alchemy “) in convolutional networks. The principles behind some important parameter tuning techniques are explained. Tuning techniques include gradient descent, learning rate, activation function, network parameter initialization, batch normalization, data enhancement, visual training process analysis, fine-tune, etc. Many tuning techniques are also applicable to other networks. After completing this course, participants can call themselves “alchemist”. .
5-1 adagrad_adam
5-2 Activating a function to a callback Technique (1)
5-3 Activating a function to a callback Technique (2)
Tensorboard Combat (1)
Tensorboard (2)
5-6 fine – most cerebral sci-film – in actual combat
5-7 activation -, initializer – optimizer – in actual combat
5-8 Use of image enhancement API
5-9 Image enhancement for actual combat
Batch 5-10 Normalized Actual combat (1)
Batch 5-11 Normalized Combat (2)
Chapter 6 image style conversion
This course is an application course of convolutional neural network, using a pre-trained VGG model to achieve image style conversion algorithm. The knowledge points of this course include feature extraction using convolutional neural network, definition of content feature and style feature, and image reconstruction method. In addition to the basic image style conversion algorithm, this course also introduces two other improved version of the style conversion algorithm. .
Application of 6-1 convolutional neural network
6-2 Capabilities of convolutional neural networks
6-3 Image style conversion V1 algorithm
6-4 VGG16 pre-training model format
6-5 VGG16 pretraining model reading function package
6-6 VGG16 model building and loading class encapsulation
6-7 Image style conversion algorithm defines input and call VGG-NET
6-8 Image style conversion calculation diagram construction and loss function calculation
6-9 Image style conversion training process code implementation
6-10 Display of image style conversion effect
6-11 Image style conversion V2 algorithm
6-12 Image style conversion V3 algorithm
Chapter 7 recurrent neural networks
This course introduces recurrent neural networks. It includes the basic structure, multilayer, bidirectional, residual structure and recursive truncated gradient descent of the recurrent neural network for solving sequential problems. The common variant – long and short term memory network is explained in detail. Various application models of recurrent neural network and convolutional neural network in text classification are explained and compared, including TextRNN, TextCNN and HAN (Hierarchical attention network, introducing attention mechanism). .
7-1 sequence problem
Let’s look at the 7-2 circulating neural network
7-3 Long and short term memory network
Text Classification Model Based on LSTM (TextRNN and HAN)
7-5 Text Classification Model Based on CNN (TextCNN)
7-6 RNN and CNN fusion to solve text classification
7-7 Participle of data preprocessing
7-8 Word table generation and category table generation for data preprocessing
7-9 Practical code module analysis
7-10 Definition of hyperparameters
7-11 word list encapsulation and category encapsulation
7-12 Data set encapsulation
7-13 Calculation diagram input definition
7-14 Calculation diagram implementation
7-15 Index calculation and gradient operator implementation
7-16 Training process realization
7-17 Realization of internal structure of LSTM unit
7-18 TextCNN implementation
7-19 Summary of circulating neural networks
Chapter 8 Images generate text
This course is a joint application course of convolutional neural network and cyclic neural network. This course introduces several model variants, including Multi-modal RNN, Show and Tell, Show Attend and Tell, etc. At the end of the course, the anti-problem text generation image is described and the adversarial neural network is introduced. By the end of May 67, students should have a good understanding of the applications of convolutional neural networks and recurrent neural networks. .
8-1 image generated text problem introduced
8-2 Image generation text evaluation indicators
8-3 Encoder-Decoder framework and Beam Search algorithm to generate text
8-4 Multi-Modal RNN model
8-5 Show and Tell model
8-6 Show attend and Tell model
8-7 Bottom-up top-down Attention model
8-8 Comparison and summary of image generation text model
8-9 Data introduction, thesaurus generation
8-10 Image feature extraction (1)- Text description file parsing
8-11 Image feature extraction (2)-InceptionV3 pre-training model to extract image features
8-12 INPUT/output files with default parameter definitions
8-13 word list loading
8-14 Convert text description to ID
8-15 ImageCaptionData class encapsulation – image features read
8-16 ImageCaptionData class encapsulation – Batch data generation
8-17 Computational graph construction – Auxiliary function implementation
8-18 Computational graph construction – Picture and word embedding
8-19 Construction of computational graph – Realization of RNN structure, loss function and training operator
8-20 Training process code
8-21 Introduction of text generation image problem and summary of this lesson
Chapter 9 adversarial neural networks
This course introduces adversarial neural networks, the latest development of deep learning. It mainly includes the idea of adversarial neural network and two specific GAN networks, deep convolution Adversarial Generative network (DCGAN) and image translation model (Pix2Pix). The knowledge points involved include generator G, discriminant D, deconvolution, U-NET and so on. .
9-1 Principle of adversarial generative network
Deep Convolutional Adversarial Generative Network DCGAN(1)
9-3 deconvolution
Deep Convolutional Adversarial Generative Network DCGAN(2)
9-5 Image translation Pix2Pix
Unpaired image translation CycleGAN(1)
Unpaired image translation CycleGAN(2)
9-8 Multi-domain image translation StarGAN
9-9 Text generated image Text2Img
9-10 Confrontation generation network summary
9-11 DCGAN combat introduction
9-12 data generator implementation
9-13 DCGAN generator implementation
9-14 DCGAN discriminator implementation
9-15 DCGAN calculation diagram construction and loss function implementation
9-16 IMPLEMENTATION of DCGAN training operator
9-17 Training process realization and effect demonstration
Chapter 10 Automatic machine learning network -AutoML
This course introduces the latest development of deep learning, automatic machine learning networks. Automatic machine learning uses recurrent neural networks to automatically search for network structural parameters that need to be adjusted to achieve better results than human alchemists. This course mainly explains three kinds of the latest automatic machine learning algorithms. The three algorithms are successively developed to automatically search for the optimal convolutional neural network structure in the field of image classification. .
10-1 AutoML introduced
10-2 Automatic network structure search algorithm I
10-3 Distributed training of automatic Network structure search algorithm I
10-4 Automatic Network structure search algorithm II
10-5 Automatic network structure search algorithm III
Chapter 11 Course Summary
Review the course as a whole
11-1 Course Summary
This course is over

**

     Baidu web disk link download