Advanced Machine Learning with ENCOG

In this course you will learn advanced topics related to machine learning for more accurate neural network predictive models. You will also learn different types of neural networks and their implementations using open source machine learning framework ENCOG.
Course info
Rating
(96)
Level
Advanced
Updated
Nov 27, 2013
Duration
4h 11m
Table of contents
Course Introduction
Network Tuning - Part 1
Network Tuning - Part 2
Neural Network Architectures Overview
Feed Forward Network - Part 1
Feed Forward Network - Part 2
Feedback Networks
Course Summary
Description
Course info
Rating
(96)
Level
Advanced
Updated
Nov 27, 2013
Duration
4h 11m
Description

Are you worried about your neural network model prediction accuracy? Are you not sure about your neural network model selection for your machine learning problem? This course will introduce you to more advanced topics in machine learning. The previous introductory course, "Introduction to Machine Learning with ENCOG 3," laid out a solid foundation of machine learning and neural networks. This course will build upon that foundation for more advanced machine learning implementations. In this course, you will learn about various neural network optimization techniques to overcome the problems of underfitting and overfitting and to create more accurate predictive models. This course will also provide an overall picture of various neural network architectures and reasons for their existence. This course will be focused towards implementation of various supervised feed forward and feedback networks. During the whole course, we will be using open source machine learning framework ENCOG to implement various concepts discussed in this course. Although the implementations in this course are ENCOG-based, concepts discussed in this course are widely applicable in other frameworks or even in custom development.

About the author
About the author

Abhishek Kumar is a data science consultant, author and speaker.He holds Master’s degree from University of California, Berkeley.His focus area is machine learning & deep learning at scale.

More from the author
Doing Data Science with Python
Beginner
6h 25m
Dec 28, 2017
R Programming Fundamentals
Beginner
7h 0m
Oct 18, 2014
More courses by Abhishek Kumar
Section Introduction Transcripts
Section Introduction Transcripts

Network Tuning - Part 1
Hi this is Abhishek Kumar and this is the second module of the course. Which is on network tuning, machine learning techniques are actually tools which enable us to work on data, find important patterns and make predictions. But unless you use it properly, until then it will not give you good results and thus can diminish the confidence of realistic holders. In this regard network tuning module is very important; this module is not about learning more machine learning techniques, rather than this module is targeted towards getting the maximum out of machine learning techniques you have learned so far. This module will focus on various tuning techniques which can help you to tune the network and the training mechanisms to get the maximum accuracy of the models, and thus including yours as well as others stake holders confidence on the provided solution. We will also discuss the reasons to tune the network and then we'll try to implement these techniques with the help of ENCOG machine learning framework. Many of these techniques will also be used throughout the course for the better performance of different types of neural network based models. I have divided this module in two parts; first part will be focused towards the network size optimization, while the second part will be focused towards different training strategies to tune the network performance. Here's the outline of this module, we will start with the network tuning in general, and the scope of

Network Tuning - Part 2
Hi this is Abhishek Kumar and this is the third module of the course, which is the second part of the network tuning. In the first part we looked at the requirement of the network tuning and we talked in detail about two issues of learning process. First was the under fitting and second was over fitting. And we had discussed the network size optimization using the pruning process to address these two key issues. In this module we will take another aspect of network tuning, which is to tune the training process itself. We will look at various strategies and how ENCOG machine learning framework can help us to implement these strategies so that we can come up with a better neural network solution to solve the machine learning problem. Although ill be talking about built in strategies provided by ENCOG framework in this module, but you can develop such strategies in your custom codes also. And they can improve your machine learning implementation significantly. We will not look at different types of training algorithms here, rather we will take only one or two training algorithms and we'll fine tune the training process. We'll see how different training strategies can help to train the model properly and to avoid the problems of under fitting and over fitting in a more conclusive fashion.

Neural Network Architectures Overview
Hi this is Abhishek Kumar and this is the fourth module of the course which is on neural network architectures overview. In this module I'll give you a brief overview of different types of neural networks and why they exist in first place. We'll also look at different categories of neural networks, and then we will take few different categories and few specific implementations in subsequent modules of this course. We will cover rest of the categories in the upcoming courses.

Feed Forward Network - Part 1
Hi. This is Abhishek Kumar, and this is the fifth module, which is on Feed Forward Networks. This module is the first part of feed forward networks. In the previous module we looked at two main categories of supervised networks, which are feed forward and feedback networks. In the first part you will learn a few basic concepts of feed forward networks. We will learn about linear and non-linear problems. In this module, we will cover linear feed forward networks. Non-linear networks will be covered in the second part in the next module. We will also take various C# demos along the way to understand linear neural networks implementation.

Feed Forward Network - Part 2
Hi. This is Abhishek Kumar, and this is the sixth module, which is on Feed Forward Networks. This module is the second part of feed forward networks. In the previous module we looked at the first part of feed forward networks where we had covered the basic concept of linear versus non-linear mapping. In the previous module we took two linear networks. First was the Adaline, and second was the Linear Perceptron network. In the second part we will focus on non-linear networks, and we will take a few specific implementations of non-linear feed forward networks. We will also take a few C# demos in this module.

Feedback Networks
Hi. This is Abhishek Kumar, and this is the seventh module, which is on Feedback Networks. In the previous module we had discussed feed forward network, which was one of the two main categories of supervised networks. In this module, we will learn about another important category of supervised networks, which are feedback networks. Feedback networks are widely used in various prediction applications, especially Time Series prediction or for casting applications. We will learn some basic concepts related to feedback networks, and we'll take a few specific feedback network implementations. We will also learn to implement these networks with the help of ENCOG machine- learning framework. We will take C# demos also along the way for deeper understanding.

Course Summary
Hi. This is Abhishek Kumar, and this is the eighth and the final module of this course where I'll give you a quick summary along with a short glimpse of what you can expect in the next part of advanced machine learning with ENCOG codes.