Machine Learning Charan Puvvala About Machine Learning Hyderabad
Machine Learning Charan Puvvala
About Machine Learning Hyderabad Call out for Speakers Venue Sponsors Organizers
Agenda Define Machine Learning History Relevance Applications Classifications Problem
What is Machine Learning? Learning: “Learning is any process by which a system improves performance from experience. ” “Machine Learning is concerned with computer programs that automatically improve their performance through experience. ” - Herbert Simon Turing Award 1975 Nobel Prize 1978
Machine Learning Examples Facebook news feed ranking, with daily ∼ 100000 impressions. Computational Biology & Drug Design Crystal structure HIV protease and some potential inhibitors, PDB Code: 4 TVH
Machine Learning Examples Web Search Engines & Recommendation Engines Find recognize input, find relevant searches, predict which results are most relevant to us, return a ranked output Recommend similar products (e. g. , Flipkart, Amazon, IMDB, Netflix etc. )
Machine Learning Examples Finance Predict if an applicant is credit-worthy Detect credit card fraud Find promising trends on the stock Market Detecting fraud financial transactions
What an autonomous car sees!!
Applications
Extreme Examples The heights of machine learning come when you see advertisements like this on the page. Suggesting you to dating sites or getting consultation for your pregnant wife This is done by the behaviour of you browsing on the internet. You are being watched all the time!! Those incognito windows don’t work enough.
Why Machine Learning? Learning and writing an algorithm It's easy for human brain but it is tough for a machine. It takes some time and good amount of training data for machine to accurately classify objects Implementation and automation This is easy for a machine. Once learnt a machine can process one million images without any fatigue whereas human brain can’t That’s why ML with Big Data is a deadly combination
Components of Machine Learning All Machine Learning algorithms have three parts a. The Output b. The Objective Function or Performance Matrix c. The Input Email Spam Classification a. The Output: Categorize email messages as spam or legitimate. b. Objective Function: Percentage of email messages correctly classified. c. The Input: Database of emails, some with human-given labels
How does one learn? How did you learn as a child? How did you teach your pet what to do? ?
How did you do that? You look at the things around you, 1. Compare them, 2. Arrange them according to similarity and then gain some insights (groups of similar items, odd ones, somehow important ones …) For this, you do not need to know the “right“ answer; it might even be difficult to define the “right” answer
How did you do that? Remarks: – 1. Do I have enough examples to understand what similar means? 2. Do I consider the right things (the right features) when I say two things are similar? 3. How do I know how many groups you want? By comparing
How do you teach pets?
Types of learning Generalization ● You have a set of examples ● Infer rules from this set and apply on new ones Supervised Learning Comparison ● You have examples but the not the correct answer ● Use metrics of similarity and compare them Unsupervised Learning Feedback ● You don’t have any examples and you don’t know what to do ● But you do know you will be rewarded when you perform good. Reinforcement Learning
Pop Quiz… Customer segmentation (Different types of customers) Face recognition (Identify the person in the photo) Spam filter (Is this email spam) Recommendation system (Which products will I buy) Identifying handwritten notes (Identify words and characters)
Problems
Solutions There are literally thousands of Machine Learning algorithms – it is impossible to know and understand them all. Selection of Machine Learning algorithms and families Decision trees Naive Bayes K-nearest neighbour (KNN) K-means Perceptron Artificial Neural Networks (ANN) Independent Component Analysis (ICA) Unsupervised Neural network models (“Restricted Boltzmann machines“) Non-negative matrix factorisation (NMF) Deep belief networks Iso. Map Random Forests Linear Regression Association analysis Ordinary least squares (OLS) Hidden Markov Model Penalised regression Kernel Approximation Principal Component Analysis (PCA) Mean. Shift
Sci. Kit learn (Machine Learning library in Python)
Next? Next talks? Hackers? Space sponsors?
- Slides: 23
