Deep learning algorithms are revolutionizing data science industry and disrupting several domains. From computer vision applications to natural language processing (NLP) use cases - every field is benefitting from use of Deep Learning models.
This course starts by assuming no knowledge about Neural Networks and Deep Learning and introduces these subjects to the student one by one.
The course helps you build a deep as well as intuitive understanding of what is Deep Learning, where can Deep Learning Models be applied and then helps you solve several real life problems using Keras and PyTorch frameworks.
Convolution Neural Networks (CNN)
Recurrent Neural Networks (RNNs)
Advanced Sequence Models including LSTM & GRU
These deep learning algorithms are powered by techniques like Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short Term Memory (LSTM), etc.
If you’re a beginner in deep learning, or you are looking to hone your existing deep learning and neural network skills, you must have asked these questions:
What is deep learning?
What are the different types of neural networks?
What part do these neural networks play in the deep learning space?
When should you choose convolutional neural networks (CNN) over recurrent neural networks (RNN) - and vice versa?
How can you build deep learning models in Python?
What is forward propagation? And what is backpropagation?
What are the different loss functions in deep learning?
What are the different deep learning frameworks?
TensorFlow vs. PyTorch vs. Keras - which deep learning framework should you choose?
The course is designed to answer all the above questions in an applied first methodology.
- Hardware for Computations in Deep Learning FREE PREVIEW
- Setting up your system
- Introduction to Google Colab FREE PREVIEW
- Understanding Google Colab Interface FREE PREVIEW
- Pre-requisites for Deep Learning
- Perceptron FREE PREVIEW
- Quiz - Perceptron
- Weights in Perceptron FREE PREVIEW
- Quiz - Weights in Perceptron
- Multi Layer Perceptron FREE PREVIEW
- Quiz - Multi Layer Perceptron
- Visualizing the neural network FREE PREVIEW
- Understanding Decision Boundary
- Quiz - Visualizing the neural network
- Forward and Backward Prop Intuition
- Quiz - Forward and Backward Prop Intuition
- Gradient Descent Algorithm
- Quiz - Gradient Descent Algorithm
- Understanding Forward Propagation Mathematically
- Quiz - Understanding Forward Propagation Mathematically
- Understanding Backward Propagation Mathematically
- Quiz - Understanding Backward Propagation Mathematically
- Backward Propagation: Matrix Form
- Why Numpy?
- Neural Network From scratch Using Numpy
- Forward Propagation (using Numpy)
- Backward Propagation (using Numpy)
- Training network (using Numpy)
- Why do we need activation functions?
- Quiz - Why do need activation functions
- Linear Activation Function
- Quiz - Linear Activation Function
- Sigmoid and tanh
- Quiz - Sigmoid and tanh
- ReLU and Leaky ReLU
- Quiz - ReLU and LeakyReLU
- Quiz - Softmax
- Tips to selecting right Activation Function
- Variants of Gradient Descent
- Quiz - Variants of Gradient Descent
- Problems with Gradient Descent
- Quiz - Problems with Gradient Descent
- Quiz - RMSPro
- Introduction to loss function
- Quiz - Introduction to Loss Function
- Binary and Categorical Cross entropy / log loss
- Quiz - Binary and Categorical cross entropy / log loss
- Overview of Deep Learning Frameworks
- Quiz - Overview of deep learning frameworks
- Understanding important Kears modules
- Understanding the problem statement: Loan Prediction
- Data Preprocessing: Loan Prediction
- Quiz - Data Preprocessing: Loan Prediction
- Steps to solve Loan Prediction Challenge
- Loading loan prediction dataset
- Defining the Model Architecture for loan prediction problem
- Training and Evaluating model on Loan Prediction Challenge
- Quiz - Training and Evaluating model on Loan Prediction Challenge
- Assignment: Big Mart Sales Prediction
- Functional API for Deep Learning
- Quiz - Functional API for Deep Learning
- Solving Loan Prediction Using Functional API in Keras
- Building a custom Model Using Functional API in keras
- Quiz - Building a Model Using Functional API
- How are images stored?
- Quiz - How are images stored
- Different Image Formats
- Quiz - Different Image formats
- Reading and stacking Images
- Converting images into different formats
- Quiz - Converting image into different formats
- Extracting edges from images
- Quiz - Extracting edges from images
- Implementation: Extracting edges from images
- Quiz - Extracting edges implementation
- Project: Image Classification "Emergency Vs Non-Emergency Vehicle"
- Notebook: Neural Network in Keras and Hyperparameter Tuning
- Neural Network in Keras
- Hyperparameter Tuning for MLP in Keras
- Assignment: Gender Classification
- Early stopping
- Early stopping: Implementation
- Quiz - Early stopping: Implementation
- Dropout: Implementation
- Quiz - Dropout: Implementation
- Vanishing and Exploding gradients
- Vanishing and Exploding gradients: Implementation
- Quiz - Vanishing and exploding gradient: Implementation
- Weights Initialization Techniques
- Implementing different weight initializing techniques
- BatchNorm: Implementation
- Advantages of Batch Normalization
- Quiz - Advantages of Batch Normalization
- Image Augmentation on Emergency-non emergency dataset
- Image Augmentation Techniques
- Image Augmentation Techniques: Implementation
- Image Generator and Fit Generator
- Assignment: Gender Classification
- Model Checkpointing
- Implementing model checkpointing
- Why do we need CNN?
- Quiz - Why do we Need CNN
- How filters work?
- Quiz - How filters work?
- Filters in CNN
- Quiz - Filters in CNN
- Parameter Sharing and Local Connectivity in CNN
- Quiz - Parameter Sharing and Local Connectivity
- CNN Architecture
- Quiz - CNN Architecture
- Quiz - Pooling
- CNN Forward Propagation
- Quiz - CNN Forward Propagation
- CNN Backward Propagation
- Quiz - CNN Backward propagation
- CNN Backprop : Matrix Form (Optional)
- Convolutional Neural Network in Keras
- Quiz - CNN in keras
- Hyperparameter Tuning for CNN in keras
- Quiz - Hyperparameter Tuning for CNN in keras
- Assignment: Gender Classification
- Model Checkpointing in CNN
- Introduction to Transfer Learning FREE PREVIEW
- How to select right pre-trained model?
- Exercise : How to select right pre-trained model?
- Steps to solve emergency vs non emergency vehicle classification challenge using transfer learning
- Exercise : Steps to build the model using transfer learning
- Solving the challenge using transfer learning (Notebook)
- Solving the challenge using Transfer Learning (Part I)
- Exercise: Solving the challenge using Transfer Learning
- Solving the challenge using Transfer Learning (Part II)
- Exercise: Solving the challenge using Transfer Learning -II
- Solving the challenge using Transfer Learning (Part III)
- Exercise : Solving the challenge using transfer learning - III
- Different fine tuning techniques
- Transfer Learning: Using the architecture of the pre-trained model
- Transfer Learning: Freezing some layers and training others
- Advantages of transfer learning
- Assignment: Gender Classification
- Introduction to Neural Network Visualization
- Quiz - Introduction to Neural Network Visualization
- How can we Interpret a Neural Network?
- Setting up the System
- Attempt 1: Understand the model architecture
- Quiz - Attempt 1: Understand the model architecture
- Attempt 2: Visualize the filters / weights
- Quiz - Attempt 2: Visualize the filters / weights
- Attempt 3: Extract the output of intermediate layers
- Quiz - Attempt 3: Extract the output of intermediate layers
- Attempt 4: Locate important parts of the image
- Quiz - Attempt 4: Locate important parts of the image
- Object Detection, segmentation, image generation
- Quiz - Object Detection, Segmentation, image generation
- Sequential Modeling
- Quiz - Sequential Modeling
- Getting Started with Text Data
- Quiz - Getting Started with Text Data
- Introduction to Text Preprocessing
- Quiz - Intro to Text-Preprocessing
- Regular Expressions in Action - I
- Regular Expressions in Action - II
- Regular Expressions in Action - III
- Quiz - Regular Expressions in Action
- Understanding Text Representation
- Introduction to One-Hot-Encoding
- Quiz - Introduction to One-Hot-Encoding
- Implementation: One-Hot-Encoding
- Limitations of One-Hot-Encoded Vectors
- Quiz - Limitations of One-Hot-Encoded Vectors
- Word Embeddings
- Implementation: Word Embeddings
- Quiz - Word Embeddings
- Journey So Far
- Introduction to RNN
- Quiz - Introduction to RNN
- Forward Propagation in RNN
- Quiz - Forward Propagation in RNN
- Backward Propagation Through Time - Part 1
- Backward Propagation Through Time - Part 2
- Backward Propagation Through Time - Part 3
- Quiz - BPTT
- Overview of Auto-Tagging System
- Quiz - Overview of Auto-Tagging System
- Understanding the Dataset
- Merging Different Datasets
- Dataset Preparation
- Model Building and Tags Prediction
- Quiz - Model Building and Tags Prediction
- Shortcomings of RNN
- Quiz - Shortcomings of RNN
- What is Long Short Term Memory (LSTM) Network?
- Quiz - What is Long Short Term Memory (LSTM) Network?
- What is Gated Recurrent Unit (GRU) Network?
- Quiz - What is Gated Recurrent Unit (GRU) Network?
- Solving Auto-Tagging Problem using LSTM and GRU
- How to use CNN for Text Data?
- Quiz - CNN for Text Data
- Solving Auto-Tagging Problem using CNN
- Assignment: Identify the Sentiments
- Overview of Web Traffic Forecasting Problem
- Quiz - Overview of Web Traffic Forecasting Problem
- Data Exploration and Pre-processing
- Quiz - Data Exploration and Pre-processing
- Model Building and Forecasting
- Introduction to Audio Data
- Understanding the Audio Classification Problem
- Audio Data Preparation
- Audio Classification using Time Domain Features
- Audio Classification using Spectrogram features
- Assignment: Urban Sound Classification
- Recap of Previous Modules
- Introduction to Unsupervised Learning
- Quiz - Introduction to Unsupervised Learning
- How to Solve Unsupervised Learning Problems?
- Introduction to Autoencoders
- Quiz - Introduction to Autoencoders
- Photo Gallery Organizing using Autoencoders
- Dataset: Photo Gallery Organization
- Implementation: Photo Gallery Organization
- Unsupervised Deep Learning - Problems and Research
- Assignment: Image Denoising
- Image Denoising Dataset
- Overview of the Module
- Introduction to PyTorch and tensors
- Mathematical and matrix operations in PyTorch - Part I
- Mathematical and matrix operations in PyTorch - Part II
- Mathematical and matrix operations in PyTorch - Part III
- Neural Network from scratch in PyTorch - Part I
- Neural Network from scratch in PyTorch - Part II
- Neural Network from scratch in PyTorch - Part III
- Important PyTorch modules (Autograd, nn, optim, etc.)
- Solving image classification problem using MLP in PyTorch - Part I
- Solving image classification problem using MLP in PyTorch - Part II
- Solving image classification problem using MLP in PyTorch - Part III
- Solving image classification problem using MLP in PyTorch - Part IV
- Convolutional Neural Network in PyTorch
- Hyperparameter Tuning of CNN
- CNN Improvements
- Transfer Learning in PyTorch - Part I
- Transfer Learning in PyTorch - Part II
- Working with text data in pytorch
- Text preprocessing in pytorch
- Building an RNN model in PyTorch
- Evaluating RNN model
- Building and evaluating LSTM in PyTorch
- Autoencoders in PyTorch
- Where to go from here?
Aishwarya is currently working as a Data Scientist at Analytics Vidhya. She is one of the primary content curators and an instructor for Analytics Vidhya’s most popular course – Applied Machine Learning. She is also an avid reader and blogger who loves exploring the endless world of data science and artificial intelligence. She has written over 70 articles in recent years on various machine learning and deep learning topics and applications.
Prateek is a Data Scientist at Analytics Vidhya. He has a multidisciplinary academic background and rich experience in BFSI and E-Learning industries. Prateek's strengths include expertise in Natural Language Processing (NLP) and Machine Learning. He is well versed in Python, R and most of the libraries and frameworks around machine learning and NLP. He has taken various trainings around NLP and Data Science and he is also a course instructor and content creator at Analytics Vidhya.
Pulkit is a Data Scientist at Analytics Vidhya. His research area lies in the field of Computer Vision and Deep Learning. He has been working on various projects related to images and videos for the past few years. He is comfortable with Python, Keras, PyTorch and has done multiple projects using these frameworks and tools. Some of his key projects include Crowd Counting, Estimating the Screen Time in videos, Object Detection, and Image Segmentation. He is one of the primary content curators for Analytics Vidhya’s courses, such as the Computer Vision using Deep Learning and Applied Machine Learning. He is also an avid blogger and has written multiple detailed and in-depth guides on various computer vision topics and applications, ranging from Image Classification to Object Detection and Image Segmentation.
Who should take the Fundamentals of Deep Learning course?
This course is for beginners in deep learning. You don’t need to know anything about neural networks - we’ll cover all of that inside the course!
How long would I have access to the “Fundamentals of Deep Learning” course?
Once you register, you will have 6 months to complete the course.
How much effort will this course take?
We suggest spending 8-10 hours per week to complete the “Fundamentals of Deep Learning” course and apply your learning on real-world projects. The time taken on projects varies from person to person.
What kind of deep learning projects can I take up?
We have provided plenty of real-world deep learning projects in the course, including object detection, image classification, among others.
What is the fee for this course?
The fee for this course is Rs. 7,999.
Can I download videos from this course?
We regularly update the "Fundamentals of Deep Learning" course and hence do not allow for videos to be downloaded.
Which programming language is used in this course?
Deep learning using Python - that’s the idea behind this course. Python is the most popular language for deep learning and you’ll see why when you take the course!
Introduction to Deep Learning
This module will introduce you to the world of Deep Learning. You will learn various applications and use cases of deep learning in the real world.
Introduction to Perceptron and Multi Layer Perceptron (Artificial Neural Network)
Then you will understand what multi layer perceptron, how they work and you will also visualize a neural network using TensorFlow playground.
Understanding Forward and Backward propagation (Along with its maths)
We will dive in mathematics behind forward and backward propagation. We will see how gradient descent is helping to update weights and biases of neural networks.
Implementing Neural Network from scratch (NumPy)
You will implement your first neural network from scratch using NumPy. You will code all the concepts of forward and backward propagation in Python.
Discuss why we need activation functions , several types of activation functions, and when should we use which one?
Type of Optimizers
We will discuss several optimizers like Momentum, RMSProp, Adam. You will also learn how you can implement these optimizers from scratch.
Loss Functions for Neural Network
In this module, you will learn several types of loss functions like Mean-Squared-Error, Binary-Cross-Entropy, Categorical- Cross-Entropy and others. We will also discuss use cases of these loss functions in different scenarios.
Introduction to Keras
You will get an overview of different deep learning frameworks like PyTorch, TensorFlow, and Keras. We will discuss various steps of model building using Keras.
Project: Solving Loan Prediction problem using Neural Network in Keras
You will build your first deep learning model on structured dataset. We will solve the “Loan Prediction problem” using neural networks in Keras.
Getting started with Image Data
This module will teach you the basics about images, how you can deal with them. You will learn how to read single as well as multiple images and then perform pre-processing on them like cropping, rotating, detecting edges, etc.
Project: Solving Image Classification problem using Neural Network (Emergency vs Non-Emergency)
You will solve an image classification problem of classifying images of vehicles as emergency or non-emergency using neural networks. You will implement this project in Keras. You will also learn how to tune hyperparameters of a neural network.
Improving your Deep Learning Model
In this module, you will learn different techniques that can be used to improve your deep learning model. We will cover techniques like Early Stopping, Dropout, Regularization, Batch Normalization, Image Augmentation etc.
Introduction to Convolutional Neural Networks (CNNs)
In this module, you will understand one of the most famous deep learning architecture convolutional neural networks (CNNs) for image data. You will learn how they works, different components of CNN and architecture
Project: Solving Image Classification problem using CNN (Emergency vs Non-Emergency)
You will solve emergency vs non-emergency vehicle classification problems using CNN and compare results with the previous deep learning model (MLP). You will also learn different hyperparameter tuning techniques for CNN.
Introduction to Transfer Learning
In this module, you will learn what is transfer learning, pre-trained models, how you should select a pre-trained model depending on the problem at hand. You will also use transfer learning to solve “Emergency vs Non-emergency” vehicle classification problem.
Visualizing Neural Networks
We will discuss several techniques to visualize your deep learning model and try to interpret reasons behind model performance.
Horizon of Deep Learning
We will explore different use cases of deep learning other than image classification like object detection, Image Segmentation, Image Captioning, Time series, text classification, audio processing and video data.
Working with Text Data
In this module the focus will be handling text data. You will learn essential text pre-processing techniques and vector representation of text.
Getting started with Recurrent Neural Network (RNN)
RNN is a special neural network architecture that is used to process sequence based information. We will also discuss forward and backward propagation methods of RNN.
Advance Sequence Models: LSTM and GRU
There are certain drawbacks of RNN which are addressed by an advanced sequence modeling architectures such as LSTM and GRU. You will learn about LSTM and GRU architectures and they function.
Project: Automatic Tagging System for Text Data
In this project, we will apply the learnings of Sequence Modeling to build a automatic tagging model for text data.
Project: Time Series Forecasting
In this project, we will learn to apply deep learning sequence models(LSTM) on time series data and do forecasting.
Project: Audio Classification
After working on image and text data, in this project, you will work with audio data. You will also learn to create features from raw audio and then use the features to build an audio classification model
Unsupervised Deep Learning
Supervised Learning is what is covered in the previous modules. Unsupervised learning algorithms derive insights directly from the data itself without labels, and work as summarizing the data or grouping it, so that we can use these insights to make data driven decisions. Here, we will understand Unsupervised Deep Learning and showcase its use for image categorization without using labels.
Introduction to PyTorch
Till now we have worked with Keras to solve all the projects of image classification, text classification, and others. Keras will solve almost 95% of your projects but when we move to advanced projects like object detection, image segmentation, it gets difficult to solve them using keras. Here, we will introduce you to another deep learning framework PyTorch and installation steps of PyTorch.
Build Deep Learning Models using PyTorch
In this module, we will build MLP, CNN and RNN models using PyTorch for various challenges like Image classification, Text Classification, Time Series and audio classification.