machine learning introduction

38
Machine Learning Pranav Prakash

Upload: pranav-prakash

Post on 07-Apr-2017

32 views

Category:

Technology


1 download

TRANSCRIPT

Page 1: Machine Learning Introduction

Machine Learning Pranav Prakash

Page 2: Machine Learning Introduction

Who am I?❖ Pranav Prakash

❖ VP Engg, Octo.ai

❖ Analytics, Recommendations, Predictions

❖ Co-Founder, Solidry

❖ 3D Printing, Computer Vision

❖ Engineering @LinkedIn SlideShare

❖ Search, Recommendations, Content Analysis

Page 3: Machine Learning Introduction

Our Agenda

❖ Artificial Intelligence

❖ Fundamental Ideas of Artificial Intelligence

❖ Machine Learning

❖ Categories

❖ Techniques

❖ Real Life Examples

Page 4: Machine Learning Introduction

Survey

❖ Artificial Intelligence

❖ Machine Learning

❖ Deep Learning

A.I.M.L.

D.L.

Page 5: Machine Learning Introduction
Page 6: Machine Learning Introduction

Machine Learning Trend

Source: Google Trends

Page 7: Machine Learning Introduction

Machine Learning vs Artificial Intelligence

Source: Google Trends

Page 8: Machine Learning Introduction

vs. Deep Learning

Source: Google Trends

Page 9: Machine Learning Introduction

Predict the Future

Page 10: Machine Learning Introduction

Fundamental Idea: Search

❖ Idea of Artificial Intelligence came in 1950s-60s

❖ Intelligence = “Intelligent Search” amongst the possible solutions

❖ Popular algo - Dijkstra

Image: https://www.cs.bham.ac.uk/~mdr/teaching/modules04/java2/DijkstraAlgo.html

Page 11: Machine Learning Introduction

Fundamental Idea: Artificial

❖ A - Star search

❖ f(n) = g(n) + h(n)

Page 12: Machine Learning Introduction

Fundamental Idea: Artificial

❖ A - Star search

❖ f(n) = g(n) + h(n)❖ Domain Knowledge❖ Explicitly Programmed❖ Artificial Intelligence

Page 13: Machine Learning Introduction

Fundamental Idea: Artificial❖ h(n, g) = Euclidean Distance

❖ h(n, g) = Diagonal Distance

❖ h(n, g) = Manhattan Distance

Image: http://theory.stanford.edu/~amitp/GameProgramming/Heuristics.html

Page 14: Machine Learning Introduction

Search Revisited

Start State

Goal State

❖ Start with the best state❖ Find the next best state

(Refinement)❖ Repeat until convergence

Page 15: Machine Learning Introduction

Fundamental Idea: Optimisation

❖ Powerful

❖ Late 1990s

❖ Start with a guess

❖ Refine the guess until convergence

Start Guess

Goal State

Page 16: Machine Learning Introduction

Fundamental Idea: Optimisation

Page 17: Machine Learning Introduction

Learning a new flower❖ I learned a new flower

❖ Experience

❖ Seeing a flower

❖ Measure:

❖ Errors in recognising flower

❖ Task

❖ See a flower and recognise it

Page 18: Machine Learning Introduction

Learning (Training)

Tag Apple Fruit Apple Corporation Peach

Color Red White Red

Type Fruit Logo Fruit

Shape Oval Cut Oval Round

Features

Page 19: Machine Learning Introduction

Typical Workflow

Tag

Input Feature ExtractorFeatures

MLAlgo

Training

Input Feature ExtractorFeatures

Model Tag

Prediction

Page 20: Machine Learning Introduction

Formal Machine Learning

A computer program is said to learnfrom experience (E)with some class of tasks (T)and a performance measure (P)if its performance at tasks in T as measured by P improves with E

Page 21: Machine Learning Introduction

Categories of Learning

❖ Supervised

❖ Unsupervised

❖ Semi-Supervised

❖ Reinforcement

Page 22: Machine Learning Introduction

Supervised Learning

❖ We are given a data set and already know what our correct output should look like, having the idea that there is a relationship between the input and the output

❖ Regression

❖ Classification

❖ Ex: Anti Spam

Page 23: Machine Learning Introduction

Unsupervised Learning❖ Allows us to approach problems with little or no idea

what our results should look like.

❖ Can derive structure from data where we don’t necessarily know the effects of the variables

❖ Clustering

❖ Ex: Photo Tagging to find out individual faces

Page 24: Machine Learning Introduction

Semi-Supervised Learning

❖ Small labelled data + Huge unlabelled data

❖ Labelled data = $$$$ + 🕰 🕰 🕰 🕰

❖ 400hrs of annotated data for 1hr of speech

❖ Ex: Infant word mapping (https://www.linkedin.com/pulse/from-baby-crying-machine-learning-james-mao)

Source: http://hnk.ffzg.hr/bibl/lrec2006/pdf/66_pdf.pdf

Page 25: Machine Learning Introduction

Reinforcement Learning

❖ Learn from self behaviour.

❖ Based on feedback from environment (Reward, Observation)

❖ Adaptive

Page 26: Machine Learning Introduction

Techniques

❖ Classification

❖ Predict class from observations

❖ Clustering

❖ Group observations into “meaningful” groups

❖ Regression (Prediction)

❖ Predict value from observations

Page 27: Machine Learning Introduction

Classification❖ Classify a document into a “predefined” category

❖ Document = Text, Image, Sound etc

❖ Popular Algorithms

❖ Naive Bayes

❖ Logistic Regression

❖ Decision Trees

❖ SVM

Page 28: Machine Learning Introduction

Clustering❖ Task of grouping a set of items in such a way that items

in same group are more similar to each other

❖ Objects are not predefined

❖ Popular Algorithms

❖ K-Means

❖ EM Clustering

❖ Affinity Propagation

Page 29: Machine Learning Introduction

Regression

❖ Prediction of a quantity given past values

❖ Popular Algorithms

❖ Linear Regression

❖ Logistic Regression

Page 30: Machine Learning Introduction

Real Life Examples

❖ Recommender Systems

❖ Learn to Rank

❖ Sentiment Analysis

❖ Object Recognition

Page 31: Machine Learning Introduction

Recommender Systems❖ Collaborative Filtering

❖ Amazon Machine Learning Library

❖ Edge Cases

Page 32: Machine Learning Introduction

Learn To Rank (LTR)❖ Used by search engines to rank

results

❖ LTR In Lucene

❖ solr-ltr, lucene-ltr, nlp4l

Page 33: Machine Learning Introduction

Sentiment Analysis

Page 34: Machine Learning Introduction

Bonanza

Page 35: Machine Learning Introduction

Object Recognition

Page 36: Machine Learning Introduction

Toolkits❖ General Purpose:

❖ Apache Mahout

❖ Apache Spark MLLib

❖ Tensor Flow

❖ FBLearner Flow

❖ Cloud

❖ IBM Watson

❖ AWS ML

❖ Azure ML

❖ H20

Page 37: Machine Learning Introduction

Advise for applying M.L.

❖ Analytics

❖ Clean data

❖ Choice of library/framework

❖ Hosted vs Managed

❖ Deployment of Models

Page 38: Machine Learning Introduction

Resources❖ https://github.com/josephmisiti/awesome-machine-learning❖ https://www.coursera.org/learn/machine-learning/home/welcome❖ https://www.slideshare.net/GaneshVenkataraman3/learn-to-rank-using-

machine-learning