Date post: | 15-Apr-2017 |
Category: |
Business |
Upload: | gopinath-chidambaram |
View: | 199 times |
Download: | 0 times |
05/03/20232
© KPIT Technologies Limited
Contents – Machine Learning – Rise of the Brain Market overview / analysis Technology process Applications ML Landscape ML Overview Why, What and How of ML Need for ML/ Value additions Types of ML CNN Advancements of ML
05/03/20233
© KPIT Technologies Limited
Demand Market Forecast
Similar numbers (CAGR-130%) forecast 2025.
05/03/20236
© KPIT Technologies Limited
Traditional Programming
Machine Learning
ComputerData
ProgramOutput
ComputerData
OutputProgram
Machine Learning - programming
05/03/20237
© KPIT Technologies Limited
Applications Web search Finance : Credit scoring,
fraud detection E-commerce Robotics Information extraction Social networks Debugging Space exploration Computational biology Retail: Market basket
analysis, Customer relationship management (CRM)
Manufacturing: Optimization, troubleshooting
Medicine: Medical diagnosis
Telecommunications: Quality of service optimization
Bioinformatics: Motifs, alignment
Example formation (feature and label
extraction)
Modeling
Evaluation
05/03/20231
1
© KPIT Technologies Limited
ML Overview “A breakthrough in machine learning
would be worth ten Microsoft” (Bill Gates, Chairman, Microsoft)
Automating automation
Getting computers to program themselves
Let the data do the work instead!
▫Learn from past experiences
▫Improve the performances of intelligent programs
Classifier
New data
Presence or absence
Database1’st data Absence2’nd data Presence… …
Training
05/03/20231
2
© KPIT Technologies Limited
AI Approach Reasoning with
Knowledge▫Knowledge base▫Reasoning
Traditional Approaches▫Handcrafted
knowledge base▫Complex reasoning
process▫Disadvantages
Knowledge acquisition bottleneck
Best move -New matrixOpponent’s
playing his move
Matrix representing the current board
Searching and
evaluating
Example: Chess Program
Example in retail: Customer transactions to consumer behavior:
05/03/20231
3
© KPIT Technologies Limited
Why “Learn”? Machine learning is programming
computers to optimize a performance criterion using example data or past experience.
There is no need to “learn” to calculate payroll
Learning is used when:▫ Human expertise does not exist
(navigating on Mars),▫ Humans are unable to explain their
expertise (speech recognition)▫ Solution changes in time (routing
on a computer network)▫ Solution needs to be adapted to
particular cases (user biometrics)
05/03/20231
4
© KPIT Technologies Limited
Feature Engineering Representation of the Real World
Data▫Features: data’s attributes
which may be useful in prediction
Feature Transformation and Selection▫Select a subset of the
features▫Construct new features,
e.g. Discretization of real value features Combinations of existing features
Post Processing to Fit the Classifier▫Does not change the
nature`
Intelligent Programs Value Functions
– Input: features– Output: value
Classifiers– Input: features– Output: a single
decision Sequence Labeling
– Input: sequence of features
– Output: sequence of decisions
Key issue in typical classifier, the results depends on feature engineering
05/03/20231
5
© KPIT Technologies Limited
What is Machine Learning? Machine Learning
▫ Study of algorithms that improve their performance at some task with experience
Optimize a performance criterion using example data or past experience.
Role of Statistics: Inference from a sample
Role of Computer science: Efficient algorithms to▫ Solve the optimization
problem▫ Representing and evaluating
the model for inference
Example: Credit scoring
Differentiating between low-risk and high-risk customers from their income and savings
Discriminant: IF income > θ1 AND savings > θ2 THEN low-risk
ELSE high-risk
05/03/20231
6
© KPIT Technologies Limited
ML FlowPattern recognitionFace recognition: Pose, lighting, occlusion (glasses, beard), make-up, hair style Character recognition: Different handwriting styles.Speech recognition: Temporal dependency.
Use of a dictionary or the syntax of the language. Sensor fusion: Combine multiple modalities; eg, visual (lip image) and acoustic for speech
Medical diagnosis: From symptoms to illnessesWeb Advertising: Predict if a user clicks on an ad on the Internet.
05/03/20231
7
© KPIT Technologies Limited
Choosing the Training Experience
Sometimes straightforward Text classification, disease diagnosis
▫Sometimes not so straightforward Chess playing
Other Attributes▫How the training experience is
controlled by the learner?▫How the training experience
represents the situations in which the performance of the program is measured?
Evaluation methods Accuracy Precision and
recall Squared error Likelihood Posterior
probability Cost / Utility Margin Entropy K-L divergence Etc.
05/03/20231
8
© KPIT Technologies Limited
Target Function Choosing the Target
Function▫ What type of knowledge
will be learned?▫ How it will be used by the
program? Reducing the Learning
Problem▫ From the problem of
improving performance P at task T with experience E
▫ To the problem of learning some particular target functions
Solving Real World Problems What Is the Input?
- Features representing the real world data
What Is the Output?- Predictions or decisions to be made
What Is the Intelligent Program?- Types of classifiers, value functions, etc.
How to Learn from experience?- Learning algorithms
Eg: Whom to target -> What they will but -> How to retain them -> How to support them
05/03/20231
9
© KPIT Technologies Limited
Types of Learning
Supervised (inductive) learning▫ Training data includes
desired outputs Unsupervised learning
▫ Training data does not include desired outputs
Semi-supervised learning▫ Training data includes a few
desired outputs Reinforcement learning
▫ Rewards from sequence of actions
05/03/20232
0
© KPIT Technologies Limited
Supervised Learning: Uses
• Prediction of future cases: Use the rule to predict the output for future inputs
• Knowledge extraction: The rule is easy to understand
• Compression: The rule is simpler than the data it explains
• Outlier detection: Exceptions that are not covered by the rule, e.g., fraud
Example: decision trees tools that create rules
Unsupervised Learning
Learning “what normally happens”
No output Clustering: Grouping similar
instances Other applications:
Summarization, Association Analysis
Example applications– Customer segmentation in CRM– Image compression: Color
quantization– Bioinformatics: Learning motifs
05/03/20232
3
© KPIT Technologies Limited
Key points CNN1. Visual scenes are generally, hierarchically organizedEg: input image -> primitive feature -> object parts -> objectForest image -> oriented edges -> bark, leaves -> trees2. Hard to train
over fittinglocal optima
3. Image statistics development4. Low level features are local5. High level features are coarser
05/03/20232
6
© KPIT Technologies Limited
Future Technologies Big data and Cloud computing Internet of Things Artificial Intelligence
Robotics Machine learning
Advancement in ML Stereoscopic 3D learning and 2D testing Video based learning – Deep intense learning for automobile applications
https://www.gazemetrix.com/docs/products/photo-monitor
http://blueapp.io/blog/the-internet-of-things-and-machine-learning/
https://www.quora.com/What-is-the-difference-between-Data-Analytics-Data-Analysis-Data-Mining-Data-Science-Machine-Learning-and-Big-Data-1
https://www.linkedin.com/pulse/machine-learning-demand-forecasting-pricing-vishnu-mohan
http://homepages.inf.ed.ac.uk/amos/publications/Storkey2011MachineLearningMarkets.pdf
References