Dropout (and why you should always consider it) | recluze

Autoplay
Autocomplete

Previous Lesson Complete and Continue

Practical Deep Learning with Tensorflow 2.x and Keras

Introduction

About the Instructor (1:53)
Dive into Machine Learning (13:10)
Making Predictions (7:01)

A Bit of Theory

Machine Learning Pipeline (9:13)
Regression (13:01)
Binary and Multi-class Classification (14:29)
Recap and a Link to More Theory (2:43)

Installation and Setup

Environment setup for Windows (and some issues with it) (5:37)
Environment setup for Mac and Linux (5:05)

Say Hi to Keras

Data Preparation (10:11)
Training and Testing (10:32)
Working with TensorBoard for Visualization (3:17)
Using Google Colab and Drive Together (7:17)

Real World Case Study: Predicting Protein Functions

Problem Description and Data View (8:32)
Pre-Processing the Data (15:51)
Loading Data and Getting the Shapes Right (7:45)
Train, Test Split (3:11)
Shapes In-Depth (or how not to get headaches for days) (4:32)
Sequential Model (8:58)
Functional API (5:25)

Convolutional Neural Networks

Basics and Rationale (10:13)
CNN in Keras (or why Keras is better than your ML tool) (8:30)
Pooling (and why it's not that important) (4:25)
Dropout (and why you should always consider it) (3:51)

Graph-Based Models

Functional API for CNN (4:27)
Inception Module (9:36)
Residual Connections (5:08)

Finishing Touches

Saving and Loading Model Weights (6:30)
Parting Words (3:55)

Transfer Learning

Transfer Learning Basics (10:09)
Transfer Learning Code - No Pre-training (6:31)
Transfer Learning Code - With Transfer and Fine Tuning (8:58)

Teach online with

Dropout (and why you should always consider it)

Complete and Continue

Discussion

0 comments