Dynamax Feb 5th Session and Full reading list

Our next live session will be at 8am pst Feb 5th**. Youtube link at the bottom of this post.**

This next session focus will be on fully understanding this on Equation 9.1 which is shown here. This equation covers all the fundamental ideas we’ll need for this book club

The goal is fully understand every single part of the equation below and be comfortable with what it represents “philosophically” and mathematically. In particular we’ll focus on

  1. Updating beliefs over time
  2. State representation and Space Representation
  3. Distributions
  4. How I learn notation

Here is our reading list
From ProbML

  • 8.1, 8.1.1
  • 9.1, 9.2 9.2.1,
  • 29.1
  • the DynaMAX README In particular what are state space models

I’ll be referencing images from Exploring Hidden Markov Models as well

Thank you all for your answers of what you were looking for in the other topic. Your responses are used to help guide this book club!


i put together an even more detailed agenda. This will feel basic to some of you, nonetheless I intend to cover it so we can ensure everyone who’s a part of this club is prepared to learn. We don’t want to leave anyone behind

Goal - Establish a knowledge foundation to dive into the Casino HMM notebooks the session after this one, covering Bayesian thinking, notation, and computational statistics

This session will feel like a college lecture more than the future sessions.

Here’s the full agenda.

Bayesian Update

Talk through the basic idea

  1. I believe a thing
  2. I saw a thing
  3. I believe a new thing
  • Go through the classic example coin flips and Covid Tests

  • Talk about thinking in distributions

    • “We dont just have one belief, a range of beliefs”
    • AB Testing example

Models vs Estimators

  • Making a model is one thing, Estimating the parameters is another thing

  • Conjugate solvers vs MCMC estimators

    • Also not the only place you’ll see Markov Chains again
  • Why this matters for this book club

How I learn notation

  1. Learn the fundamental ideas
  2. Basic Examples
  3. Code - Its unambiguous in code
  4. Ask people
  • Introduce our Notation Dictionary

State Space models

  • As the name implies there’s two parts

    • The State
    • The Space
  • Not just one model, but a “zoo of models” as Kevin likes to say

  • State Space Use cases

  1. Smoothing - Where am I now before I got there?
  2. Filtering - Now that I got there where was o?
  3. Prediction - Where will be?

Wrap up

  • Feeling ready for Casinos and HMMs?
  • What would you like to learn next session?
1 Like

Just a heads up for anyone that is working through this reading list. If you are struggling (like I was) to understand how the joint distribution equations are derived from the DAGS then read chapter 4 where probabilistic graphs are discussed in more detail and derivations from Markov models are presented.

1 Like

Notebooks for tomorrows book club are here. Looking forward to our next live session shortly here