Probabilistic Graphical Models, HMMs using PGMPY
PGMs are generative models that are extremely useful to model stochastic processes. I shall talk about how fraud models, credit risk models can be built using Bayesian Networks. Generative models are great alternatives to deep neural networks, which cannot solve such problems. This talk focuses on Bayesian Networks, Markov Models, HMMs and their applications. Many areas of ML need to explain causality. PGMs offer nice features that enable causality explanations. This will be a hands-on workshop where attendees shall learn about basics of graphical models along with HMMs with the open source library, pgmpy for which we are contributors. HMMs are generative models that are extremely useful to model stochastic processes. This is an advanced area of ML that is helpful to most researchers and ML community who are looking for solutions in state-space problems. This workshop shall have students learn basics needed to learn about HMMs including advanced probability, generative models, markov theory and HMMs. Students shall build various interesting models using pgmpy.
Outline/Structure of the Workshop
This workshop will begin by teaching the basics of advanced probability theories, stochastic processes along with various applications. This shall be followed by markov theory and HMMs.
1. Advanced Probability Theory
2. Stochastic Processes
3. Generative Models
4. Markov Theory
5. Hidden Markov Models
6. Examples
Learning Outcome
Students shall learn what are generative models, markov theory, PGMs and Hidden Markov Models. Students shall also learn to use pgmpy, which is an open source library for modeling HMMs. The learn by examples will help students to incorporate HMMs into their own work/research.
Target Audience
Those who are familiar with Machine Learning basics
Prerequisites for Attendees
Attendees are recommended to get their laptops and know basics of Python. We propose to use refactored.ai that has in-built jupyter. All you need is a browser sign up and start using it.
Video
Links
A related topic was provided as a workshop to Freddie Mac. Here are past public talks with links still active:
schedule Submitted 2 years ago
People who liked this proposal, also liked:
-
keyboard_arrow_down
Naresh Jain / Dr. Arun Verma / Dr. Denis Bauer / Favio Vázquez / Sheamus McGovern / Drs. Tarry Singh / Dr. Tom Starke / Dr. Veena Mendiratta - Unanswered Questions - Ask the Experts!
Naresh JainFounderXnsioDr. Arun VermaSr. Quantitative Researcher & Head of Quant Solutions TeamBloomberg L.P.Dr. Denis BauerTeam Leader Transformational BioinformaticsCSIROFavio VázquezSr. Data ScientistRaken Data GroupSheamus McGovernFounderOpen Data ScienceDrs. Tarry SinghCEO, Founder & AI Neuroscience Researcherdeepkapha.aiDr. Tom StarkeCEOAAAQuantsDr. Veena MendirattaResearch LeaderNokia Bell Labsschedule 2 years ago
45 Mins
Keynote
Beginner
Through the conference, we would have heard different speaker's perspective and experience with Data Science and AI. In this closing panel, we want to step back and look at any unanswered questions that the audience would have.
-
keyboard_arrow_down
Tamaghna Basu - Killing the Password via Gesture Recognition
20 Mins
Talk
Intermediate
We are living in times where we still have to type the password. NeoEyed's gesture recognition is able to detect swipes to authenticate the user to provide access to the mobile. This is a non-trivial problem as many users can have similar gestures. It is important that the classifiers are able to detect fine changes between many users who could potentially break into the phone. We shall talk about how various classifiers are used for such a detection. We shall talk about classifiers such as one-class SVM, multi-level robust classifiers which are useful for this scenario. Our detection mechanism won us the "Paypal Award". We are now building the next generation authentication system for which we shall discuss the technologies and challenges that need to be resolved.
The talk shall be organized as:
1. The problem we are trying to solve. 3
2. What classifiers help in such a scenario. 10
3. Challenges in existing out of box methods 3.
4. What directions are helping us. 4
-
keyboard_arrow_down
Harish Kashyap - Probabilistic Graphical Models (PGMs) for Fraud Detection and Risk Analysis.
45 Mins
Talk
Advanced
PGMs are generative models that are extremely useful to model stochastic processes. I shall talk about how fraud models, credit risk models can be built using Bayesian Networks. Generative models are great alternatives to deep neural networks, which cannot solve such problems. This talk focuses on Bayesian Networks, Markov Models, HMMs and their applications. Many areas of ML need to explain causality. PGMs offer nice features that enable causality explanations.
Public Feedback