[SOLVED] CS algorithm deep learning data mining AI LECTURE 1 TERM 2:

LECTURE 1 TERM 2:
MSIN0097
Predictive Analytics
A P MOORE

PREDICTIVE SYSTEMS
How much better is prediction than it used to be?

ALPHA GO
https://deepmind.com/research/case-studies/alphago-the-story-so-far

ALPHA STAR
https://deepmind.com/blog/article/alphastar-mastering-real-time-strategy-game-starcraft-ii

ALPHA FOLD
https://deepmind.com/blog/article/alphafold-casp13

ALPHA GO TRAINING EFFICIENCY

IMAGENET PERFORMANCE
https://www.eff.org/ai/metrics Prediction Machines, page 28

AUTONOMOUS VEHICLES CHEAP CAMERAS

MACHINE LEARNING PUBLICATIONS ARXIV PAPERS PER YEAR
Source: https://bit.ly/2JU4BmE

HOW DO I GET STARTED?
Source: @random_forests https://twitter.com/random_forests/status/1178687761177554946

ICLR2020 PAPER SUBMISSIONS TRANSFORMERS, NLP, BERT
Source: @priz77 https://twitter.com/prlz77/status/1178662575900368903

INDUSTRIALISATION
NTERESSENGEMEINSCHAFT FARBENINDUSTRIE AG
(DYE INDUSTRY SYNDICATE, INC.) COMMONLY KNOWN AS IG FARBEN
https://en.wikipedia.org/wiki/Sulfonamide_(medicine)

EMPIRICAL SCIENCE

CORPORATE PUBLICATIONS TOP 10 NEURIPS 2019 CONTRIBUTORS
Source: https://twitter.com/chipro/status/1170174478304759809
Hartmann, P. and Henkel, J. (2019). The Rise of Corporate Science in AI.
Out of 1429 accepted papers, 167 (~12%) have at least one author from Google/DeepMind, same as Microsoft, Facebook, IBM, & Amazon combined.
Co-authors in the five leading machine learning conferences from 2011 to 2018 and found that research by large firms features on average one more co-author (4.3) than non-large firm papers (3.4)
[Hartmann and Henkel (2019)]

UNIVERSITY PUBLICATIONS TOP 18 NEURIPS 2019 CONTRIBUTORS
Source: https://twitter.com/chipro/status/1170174478304759809

LANGUAGE BENCHMARKS 7TH JUNE 2019
Source: https://gluebenchmark.com/leaderboard

TWO DISTINCT ERAS OF COMPUTE IN AI
Source: https://openai.com/blog/ai-and-compute

NLP PROGRESS
PERFORMANCE ON PUBLIC BENCHMARKS CAN RIVAL HUMAN PERFORMANCE
General Language Understanding Evaluation (GLUE)

MANY APPLICATIONS
Source: London: The AI Growth Capital of Europe, Cognition X, 2019

INTRODUCTION TO AI
What is it?

TYPES OF PROBLEMS

QUESTIONS WE WOULD LIKE ANSWERED
Descriptive Analytics, which use data aggregation and data mining to provide insight into the past and answer:
What has happened?
Predictive Analytics, which use statistical models and forecasts techniques to understand the future and answer:
What could happen?
Prescriptive Analytics, which use optimization & simulation algorithms to advice on possible outcomes and answer: What should we do?

BUSINESS PROBLEMS
CAC/LTVAnalysis
Cohort Analysis
FunnelAnalysis
Customer Segmentation

Past
Descriptive
What has happened?
Click Through Prediction Demand Prediction
Fraud Detection
Sentiment Analysis
Recommender Systems Fraud Detection
Predictive Lead Scoring Dynamic Pricing

Predictive
What could happen?
Prescriptive
What should we do?
Future

H YPE?
GARTNER HYPE CYCLE, 2016

ANALYTICS AND BUSINESS INTELLIGENCE GARTNER HYPE CYCLE, JULY 2019

RISE OF COMPUTING POWER
Source: David Barber, UCL

TWO DISTINCT ERAS OF COMPUTE IN AI
Source: https://openai.com/blog/ai-and-compute

IMAGENET PERFORMANCE

TRAINING TIMES

PROJECTED GROWTH OF DEEP LEARNING CHIPSETS GLOBAL REVENUE ($bn)
Source: https://on.ft.com/2Hc6mcc

EXPENDITURE ON R&D
ES (13TH LARGEST ECONOMY, 47 MILLION PEOPLE ) VS APPLE INC. (m)

GET MORE DATA

MACHINE LEARNING
Data + modelaprediction

MACHINE LEARNING DATA DRIVEN AI
Assume there is enough data to find statistical associations to solve specific tasks
Data + modelaprediction
Define how well the model solves the task and adapt the parameters to maximize performance

WAVES OF INVESTMENT THE ROLE OF INDUSTRIALIZATION
https://machinelearnings.co/winning-strategies-for-applied-ai-companies-f02cac0a6ad8

TOWER & MOAT
Source: https://blog.gardeviance.org/2014/07/tower-and-moat.html

ML JOBS
STACKOVERFLOW DEVELOPER SURVEY 2016
http://stackoverflow.com/research/developer-survey-2016

NOT JUST CATS

CHIHUAHUA OR MUFFIN?

INTRODUCTION TO AI
Using ML in business

ML IN BUSINESS
I. Increasing the number of customers
II. Serving customers better
III. Serving customers more efficiently
Source: https://www.louisdorard.com/

INCREASING THE NUMBER OF CUSTOMERS
1. Reducing customer attrition (churn) 2. Acquiring new customers
i. Lead scoring
ii. Optimizing marketing campaigns
Prediction Machines, page 32

SERVING CUSTOMERS BETTER
3. Cross-selling products
4. Optimizing products and pricing
5. Increasing engagement

SERVING CUSTOMERS EFFICIENTLY
6. 7. 8.
Predicting demand Automating tasks
Making predictive applications
i. Prioritizing things
ii. Adapting workflows
iii. Optimizing configurations
Predictive maintenance
9.

INTRODUCTION TO AI
Using ML in business functions

INTRODUCTION TO AI
Horizontal applications

MARKETING APPLICATIONS
Predicting Lifetime Value (LTV)
Wallet share estimation
Churn
Customer segmentation
Product mix
Cross selling
Recommendation algorithms
Up-selling
Channel optimization
Discount targeting
Reactivation likelihood
Adwords optimization and ad buying
Source: http://www.fast.ai/

RISK APPLICATIONS
Credit risk
Treasury or Fraud currency risk detection
Accounts Payable Recover y
Anti-money laundering
Source: http://www.fast.ai/

HUMAN RESOURCES APPLICATIONS
Resume screening
Employee churn
Training Talent recommendation management
Source: http://www.fast.ai/

INTRODUCTION TO AI
Vertical applications

HEALTHCARE APPLICATIONS
Claims review prioritization
Alerting and diagnostics from real-time patient data
Survival analysis
Medicare/medicaid fraud
Prescription compliance
Medication (dosage) effectiveness
Medical resources allocation
Physician attrition
Readmission risk

RETAIL APPLICATIONS
Price optimization
Inventor y Management (how many units)
Location of new stores
Shrinkage analytics
Product layout in stores
Warranty Analytics
Merchandizing
Market Basket Analysis
Cannibalization Next Best Offer In store traffic Analysis Analysis patterns

TRAVEL APPLICATIONS
Aircraft scheduling
Customer complaint resolution
Maintenance Tourism optimization forecasting
Seat/gate Air crew management scheduling
Dynamic pricing

OTHER DOMAINS
Life Sciences
Identifying biomarkers
Drug/chemical discovery
Analyzing study results
Identifying negative responses
Diagnostic test development
Diagnostic targeting
Predicting drug demand
Prescription adherence
Putative safety signals
Social media marketing
Image analysis Clinical trial design COGS optimization
Insurance
Hospitality
Manufacturing
Direct Marketing
Response rates
Segmentations for mailings
Reactivation likelihood
RFM
Discount targeting
Phone marketing
Email Marketing
Construction
Agriculture
Mall Operators
Education
Utilities
Claims prediction
Claims handling
Price sensitivity
Investments
Agent & branch performance
DM, product mix
Dynamic pricing
Promos/upgrades/ offers
Table management & reservations
Workforce management
Failure analysis
Quality management
Inventory management
Warranty/pricing
Contractor performance
Yield management
Tenant capacity to pay
Automated essay scoring
Optimize Distribution Network
Design issue prediction
Automation
Tenant selection
Dynamic courses
Predict Commodity Requirements

INTRODUCTION TO AI
What do we do with data?

LECTURE 1 TERM 2:
MSIN0097
Predictive Analytics
A P MOORE