Bettering Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization

About this Course

Within the second course of the Deep Studying Specialization, you’ll open the deep studying black field to know the processes that drive efficiency and generate good outcomes systematically.

By the tip, you’ll be taught one of the best practices to coach and develop check units and analyze bias/variance for constructing deep studying functions; be capable of use normal neural community methods similar to initialization, L2 and dropout regularization, hyperparameter tuning, batch normalization, and gradient checking; implement and apply quite a lot of optimization algorithms, similar to mini-batch gradient descent, Momentum, RMSprop and Adam, and verify for his or her convergence; and implement a neural community in TensorFlow. The Deep Studying Specialization is our foundational program that can assist you to perceive the capabilities, challenges, and penalties of deep studying and put together you to take part within the improvement of modern AI expertise. It gives a pathway so that you can achieve the information and expertise to use machine studying to your work, stage up your technical profession, and take the definitive step on the planet of AI.

SKILLS YOU WILL GAIN

• Tensorflow
• Deep Studying
• Mathematical Optimization
• hyperparameter tuning

Syllabus – What you’ll be taught from this course

Content material Score96%(64,946 rankings)

WEEK

1

14 hours to finish

Sensible Facets of Deep Studying

Uncover and experiment with quite a lot of totally different initialization strategies, apply L2 regularization and dropout to keep away from mannequin overfitting, then apply gradient checking to determine errors in a fraud detection mannequin.

WEEK

2

6 hours to finish

Optimization Algorithms

Develop your deep studying toolbox by including extra superior optimizations, random minibatching, and studying charge decay scheduling to hurry up your fashions.

WEEK

3

7 hours to finish

Hyperparameter Tuning, Batch Normalization and Programming Frameworks

Discover TensorFlow, a deep studying framework that permits you to construct neural networks shortly and simply, then prepare a neural community on a TensorFlow dataset.