Sie sind auf Seite 1von 17

DEEP LEARNING

Introduction
Deep learning (also known as deep structured learning or hierarchical
learning) is part of a broader family of machine learning methods based
on learning data representations, as opposed to task-specific algorithms.
Learning can be supervised, semi-supervisedor unsupervised.

Data Analysis Pipeline

Extract
Gather Clean
Featur Learn Predict
Data Data
es

2
Feature Extraction

● Useful information in data is usually hidden


● We need some mechanism to keep only the useful parts
● Example: Predict prices of houses in Bangalore

Area (in Location Rooms Wall color Price (in


sqft) crores)

1200 Whitefield 2 Blue 1.00

1900 Indiranagar 3 Grey 2.50

2400 Jayanagar 4 Yellow 3.50

1900 Banashankar 3 Red 1.50


i

800 Lavelle 1 Green 1.50


Road
3
NEURAL NETWORK

The term neural network was traditionally used to refer to a network or


circuit of neurons.[1] The modern usage of the term often refers to artificial
neural networks, which are composed of artificial neurons or nodes.

Types of Neural Networtk

1. Multilayer perceptron (MLP)


2. Convolutional neural network (CNN)
3. Recursive neural network (RNN)
4. Recurrent neural network (RNN)
5. Long short-term memory (LSTM).
Multilayer Perceptron
b1
Input: 784 (28*28)

b3
Hidden 2: 100

X W1 W2 W3

Softmax
ReLU

ReLU
y

[1x784] [784x300] [300x100] [100x10] [1x10]

Output: 10 classes

Hidden 1: 300
b2

Stochastic Gradient Descent Loss: Categorical


Cross entropy
Formulae

• ReLU: 𝑓 𝒙 = max 0, 𝒙
𝑒 𝑥𝑖
• Softmax: 𝑓 𝒙 = σ𝑛 𝑥
𝑖=1 𝑒 𝑖
• Cross Entropy: 𝑓 𝒂, 𝒃 = − σ 𝒂 ∙ log 𝒃
• SGD: 𝒘′ = 𝒘 − 𝑙 ∙ ∆𝒘, ∀𝒘
Convolution Neural Networks (CNN)

• Brain child of Yann Lecun


• Most popular deep learning architecture
• CNN is now the go-to model on every image related
problem.
• Applications:
– Face Recognition
– Object recognition

IBM RESEARCH
Recurrent Neural
Network

.
• Data is temporal/seqential
RNN – Encoder Decoder Model

Language A Language B
Question Answer
Sentence Encode Decode Paraphrase
Compound word r r Constituent
Paragraph words
... Summary
...

Sequence-to-Sequence models

IBM RESEARCH
Generative Adversarial Networks

• Generative Network:
• Producing images that are closer to real images.
• Discriminative Network:
• Determine difference between real and fake images.

Aim: Generative network that can produce images which are indistinguishable
from the real ones.

Use cases:
• Generating realistic artwork samples (video/image/audio)
• Image to Image Translation
• Image super resolution
• Image Inpainting
• Text to Image
IBM RESEARCH
Generative Adversarial Networks
Deep Learning Libraries
Three Competitive Advantages of Deep
Learning for the Company
1. Cost and time benefits.
Neural networks are trainable “brains.” You give them your company’s
information and train them to do a task, such as generating reports, and
they will use that training, new information, and their “work experience” to
adapt and improve in much the same way a human worker learns.

2. Quality and accurate results.


Using deep learning, your software robots can recognize more data and
images, understand spoken language, overcome problems, and work more
efficiently.
3. Job growth.
Utilizing deep learning for your business will save you money and time
in more ways than one. You, your company, and your employees will
benefit when you decide to take advantage of what AI has to offer.
IMPORTANCE OF DEEP LEARNING
Application of Deep Learning
Deep learning added a rapidly developing field of computer vision. With deep
learning, a lot of new applications of computer vision techniques have been
introduced.These include face recognition and indexing, photo stylization or
machine vision in self-driving cars. deep learning models. We will cover both
image and video recognition, including image classification and annotation,
object recognition and image search, various object detection techniques,
motion estimation, object tracking in video, human action recognition, and
finally image stylization, editing and new image generation.
Conclusion

1. Deep learning = learning hierarchical representations.


2. Deep learning is thriving in big data analysis , image
processing ,
Speech recognitition and natural language processing .
3. Deep learning has matured and is very promising as an
Artificial intelligence
method
THANK YOU

Das könnte Ihnen auch gefallen