Uploaded July 2026 | Updated November 2026, 2 days ago
Book: Fundamentals of Active Inference
Principles, Algorithms, and Applications of the Free Energy Principle for Engineers, by Sanjeev V. Namjoshi https://mitpress.mit.edu/9780262050951/fundamentals-of-active-inference/
Register: http://textbook-group.activeinference.institute/
Code: github.com/ActiveInferenceInstitute/fundamentals
Playlist of Textbook Group: youtube.com/playlist?list=PLNm0u2n1IwdrpoTg-9l37YIBiDrrOzTw-
Direct link to group document: coda.io/d/_dZHp_UGhVZE/Meetings_su3ToqMp
------
CHAPTERS
0:00 Intro: opening Part Two of the book
0:23 Why Chapter 6 is where hidden states become dynamic
0:56 Temporality, flow, and one observation at a time
2:16 The Part Two overview pages (124-125)
2:49 Where Part Two sits: fundamentals, core, applications
3:22 Appendix C8 on dynamical systems, and why to read it
4:37 Observations one by one rather than in batch
5:02 Object and process: what a statistical model can account for
5:56 Filtering: what is the hidden state given my observations?
6:26 How this relates to linear regression and batch fitting
8:36 Filtering as an audio metaphor: from moving average to structured models
10:21 Perceptual filtering as a partially observable Markov process
11:42 What Chapter 7 adds: action as another modelled cause
12:37 Why you optimise cognitive and active states, not observations
14:08 Flow, state transitions, and the state transition model
15:33 Why state transitions matter for the rest of the book
16:34 Chapters 6-8 continuous, Chapters 9-10 discrete
17:24 Two approximation strategies: Laplace and mean field
19:34 Equation 6.1 and the notations for derivatives
20:21 The environment's dynamics vs the agent's model of them
21:53 6.1 Laplace-encoded energy
23:21 From static hidden states to hidden states that change
24:40 Page 128: how dynamical systems are addressed
25:05 The hidden state update rule and the initial value problem
26:06 Why this is an iterative process, not a batch fit
26:54 Why this is a major juncture in the book
27:14 Modelling hidden state change from unknown signals to rich domains
28:29 Coloured noise: patterns inside the noise
30:10 Covariance structure and exploiting noise for prediction
31:05 Why neuroscience takes noise seriously
32:13 6.1, 6.2, 6.3 as the theoretical core
33:23 The three big ideas: state dynamics, Laplace, generalized coordinates
35:34 Examples 6.1 and 6.2 as the canonical implementations
36:59 Univariate, parallel multivariate, and joint multivariate
37:43 Figure 6.4: beliefs starting too high and locking on
39:30 Reading the loss trace, and what free energy does and does not tell you
41:05 Non-Bayesian alternatives: sliding windows, regressions, time series models
41:52 Q&A: where does the mean field approximation come in?
43:56 Example 6.1 in the language of precision-weighted prediction errors
44:33 Two kinds of error: observation and state dynamics
45:00 Laplace as expansion to the second derivative
46:44 Why Laplace can be run iteratively
47:48 Quadratic curvature vs a true statistical distribution
48:21 Q&A: discontinuous events and degraded information
49:05 Switching linear dynamical systems as one answer
50:30 Priors over how many latent states there are
51:14 A robotic arm example of a discontinuity
52:29 Catching a ball: why temporality changes what you can predict
54:57 Cochlear implants and learning from a degraded signal
55:58 Wrap-up/_dZHp_UGhVZE/Meetings_su3ToqMp
------
Active Inference Institute information:
Website: https://activeinference.institute/
X: https://x.com/InferenceActive
Discord: https://discord.activeinference.institute/
Donate: https://donate.activeinference.institute/
YouTube: youtube.com/c/ActiveInference
Active Inference Livestreams: https://video.activeinference.institute/
Book: Fundamentals of Active Inference
Principles, Algorithms, and Applications of the Free Energy Principle for Engineers, by Sanjeev V. Namjoshi https://mitpress.mit.edu/9780262050951/fundamentals-of-active-inference/
Register: http://textbook-group.activeinference.institute/
Code: github.com/ActiveInferenceInstitute/fundamentals
Playlist of Textbook Group: youtube.com/playlist?list=PLNm0u2n1IwdrpoTg-9l37YIBiDrrOzTw-
Direct link to group document: coda.io/d/_dZHp_UGhVZE/Meetings_su3ToqMp
------
CHAPTERS
0:00 Intro: opening Part Two of the book
0:23 Why Chapter 6 is where hidden states become dynamic
0:56 Temporality, flow, and one observation at a time
2:16 The Part Two overview pages (124-125)
2:49 Where Part Two sits: fundamentals, core, applications
3:22 Appendix C8 on dynamical systems, and why to read it
4:37 Observations one by one rather than in batch
5:02 Object and process: what a statistical model can account for
5:56 Filtering: what is the hidden state given my observations?
6:26 How this relates to linear regression and batch fitting
8:36 Filtering as an audio metaphor: from moving average to structured models
10:21 Perceptual filtering as a partially observable Markov process
11:42 What Chapter 7 adds: action as another modelled cause
12:37 Why you optimise cognitive and active states, not observations
14:08 Flow, state transitions, and the state transition model
15:33 Why state transitions matter for the rest of the book
16:34 Chapters 6-8 continuous, Chapters 9-10 discrete
17:24 Two approximation strategies: Laplace and mean field
19:34 Equation 6.1 and the notations for derivatives
20:21 The environment's dynamics vs the agent's model of them
21:53 6.1 Laplace-encoded energy
23:21 From static hidden states to hidden states that change
24:40 Page 128: how dynamical systems are addressed
25:05 The hidden state update rule and the initial value problem
26:06 Why this is an iterative process, not a batch fit
26:54 Why this is a major juncture in the book
27:14 Modelling hidden state change from unknown signals to rich domains
28:29 Coloured noise: patterns inside the noise
30:10 Covariance structure and exploiting noise for prediction
31:05 Why neuroscience takes noise seriously
32:13 6.1, 6.2, 6.3 as the theoretical core
33:23 The three big ideas: state dynamics, Laplace, generalized coordinates
35:34 Examples 6.1 and 6.2 as the canonical implementations
36:59 Univariate, parallel multivariate, and joint multivariate
37:43 Figure 6.4: beliefs starting too high and locking on
39:30 Reading the loss trace, and what free energy does and does not tell you
41:05 Non-Bayesian alternatives: sliding windows, regressions, time series models
41:52 Q&A: where does the mean field approximation come in?
43:56 Example 6.1 in the language of precision-weighted prediction errors
44:33 Two kinds of error: observation and state dynamics
45:00 Laplace as expansion to the second derivative
46:44 Why Laplace can be run iteratively
47:48 Quadratic curvature vs a true statistical distribution
48:21 Q&A: discontinuous events and degraded information
49:05 Switching linear dynamical systems as one answer
50:30 Priors over how many latent states there are
51:14 A robotic arm example of a discontinuity
52:29 Catching a ball: why temporality changes what you can predict
54:57 Cochlear implants and learning from a degraded signal
55:58 Wrap-up/_dZHp_UGhVZE/Meetings_su3ToqMp
------
Active Inference Institute information:
Website: https://activeinference.institute/
X: https://x.com/InferenceActive
Discord: https://discord.activeinference.institute/
Donate: https://donate.activeinference.institute/
YouTube: youtube.com/c/ActiveInference
Active Inference Livestreams: https://video.activeinference.institute/










