Explore the fundamentals and applications of deep learning in this 51-minute conference talk from GOTO Copenhagen 2016. Begin with the basics of deep learning, including its definition, functionality, and how to get started. Delve into commonly used architectures such as convolutional and recurrent networks. Learn about key concepts like backward pass, parameter updates, and backpropagation. Examine matrix multiplication, sigmoid functions, and neural network types including RNNs and CNNs. Discover deep learning's potential in areas like language understanding, visual comprehension, game playing, and automated programming. Gain insights into advanced topics such as hierarchical structure discovery, memory utilization, and differentiable data structures. Conclude with an overview of module networks and the exciting possibilities deep learning offers for problem-solving across various domains.
Deep Learning - What is it and What It Can Do For You
Overview
Syllabus
Intro
Super High level
Deep Learning An Interface
Deep Learning Vocabulary
Details: How to implement the backward pass
Details: How do we update the parameters
Details: What is backproplagation
Basics Review
Matrix Multiplication
Sigmoid
Slight Tangent
Recurrent Neural Networks CRNNO
Convolutional Neural Networks (CNN)
Robot Overlords
Generation
Deeper Language Understanding
Deeper Visual Understanding
Harder Game Playing
Automatically Discovering Hierarchical Structure
Learning to Use Memory
Differentiable Data Structures
Learning to Program
Learning to Execute
Module Networks
OKI I'm pumped to deep learning some problems
Taught by
GOTO Conferences
