Introduction to Deep Learning
-
- Course level: Intermediate
About Course
In this program we will learn about the basics of deep neural networks and their applications to various AI tasks. Advance your knowledge about deep neural networks through this online certificate program with cutting-edge content offered in association with Carnegie Mellon University and understand its application in industry practice.Description
Learn how to apply neural networks and deep learning techniques to develop innovative solutions.
What we offer
Introduction to Deep Learning is a course that helps you gain footing in today’s complex job landscape.
Taught by Dr. Bhiksha Raj, this course provides you a blue-print to handle real-world challenges while imbibing autonomy to take decisions.
So what is the course about?
In this program we will learn about the basics of deep neural networks and their applications to various AI tasks. By the end of the program, it is expected that students will have significant familiarity with the subject and be able to apply Deep Learning to a variety of tasks. They will also be positioned to understand much of the current literature on the topic and extend their knowledge through further study.
What’s holding you back?
The program for Deep Learning will span 10 weeks with LIVE online, collaborative lectures covering different subareas in each week. They will be combined with assignments and project work to be completed by individuals and in groups.
Lecture Topics
Introduction to Neural Networks
- A brief introduction with history
- Neural networks as a universal approximator
Concepts behind training neural networks
- The problem of training
- Convergence issues and speed of training
- Incremental updates and second order methods
- Regularization and other tricks of the trade
Convolutional Neural Networks (CNNs)
- Scanning with neural networks
- Convolutional neural networks in detail
- Training CNNs and transpose convolutions
Recurrent neural networks
- Analyzing time series data with neural nets
- Stability and convergence
- Training recurrent nets
- Connectionist temporal models
- Attention models
Labs
Lab A
- Building your own MLP from scratch
- Building your own CNN from scratch
- Building your own recurrent network
Lab B
- Training a large MLP
- Face identification and recognition
- Speech recognition
- Sequence-to-sequence conversion
Certificate of Completion
from Carnegie Mellon University
Join this online certificate program, offered by Carnegie Mellon University, to learn as well as upgrade your knowledge on artificial neural networks.
Course Curriculum
Module 1: Introduction to Neural Networks
- History & cognitive basis of neural computation
- Connectionist Machines
- Rosenblatt’s Perceptron
- Multilayer Perceptrons
- The neural net as a universal approximator
Week: 1 & 2 Assessment: Quiz and Classroom Discussion
Module 2: Training Neural Networks
- Training a Neural Network
- Perceptron learning rule
- Optimization by gradient descent
- Convergence of perceptron algorithm
- Back propagation, Calculus of back propagation
- Convergence issues in back propagation
- Second order methods
- Convergence in Neural Networks and rates of Convergence
- Loss surfaces
- Learning rates, and optimization
methods - Stochastic gradient descent
Acceleration - Overfitting and regularization
- Tricks of The Trade:
- Choosing a divergence (loss)
function - Batch normalization
- Choosing a divergence (loss)
Week: 3 & 4 Assessment: Curated Assignments, Quiz and TA Interactions
Labs
Learning to code networks:
- Building your own CNN from scratch
- Building your own MLP from scratch
- Building your own recurrent network
Module 3: Convolutional Neural Networks: Scanning for Patterns
- Convolutional Neural Networks (CNN)
- Weights as templates
- Translation invariance
- Training with shared parameters
- Arriving at the convolutional model
- Models of vision
- Neocognitron
- Mathematical details of CNNs
- Backpropagation in CNNs
- Variations in the basic model
- Some history of the ImageNet
Week: 5, 6 & 7 Assessment: Curated Assignments, Quiz and TA Interactions
Module 4: Recurrent Neural Networks: Scanning for Patterns
- Recurrent Neural Networks (RNNs)
- Modeling series
- Back propagation through time
- Bidirectional RNNs
- Stability, Exploding/vanishing gradients
- Long Short-Term Memory Units (LSTMs) and variants
- Loss functions for recurrent
networks - Sequence prediction, Sequence
To Sequence Methods & Models - Labelling Unsegmented Sequence
- Data with Recurrent Neural
Networks
Week: 8,9 & 10
Project Day / Buffer / Overflow
Assessment: Curated Assignments, Quiz and TA Interactions, Capstone Project
Labs
Using existing toolkits to solve real-world issues:
- Training a large MLP
- Face identification and recognition
- Speech recognition
- Sequence-to-sequence conversion
What Will I Learn?
- Learn directly from Carnegie Mellon University’s top faculty
- Develop intuitive skills through practice & collaboration
- Gain confidence and accelerate your career
- Build a career in a niche, growing domain
About the instructors
1 Courses
1 students