Rotating Features For Object Discovery
Published:
Structure is a useful but underleveraged inductive bias for representation learning.
Blog posts
2023
The Machine Learning Interview Checklist
Published:
Cover your bases.
2022
Higgins et al. - Towards a Definition of Disentangled Representations
Published:
Disentanglement is a concept rooted in geometric deep learning.
Where is the nature of the relationship expressed in causal models?
Published:
Graphs don’t tell about the nature of dependence, only about its (non-)existence.
AMMI 3 Notes: Geometric priors I
Published:
In the previous post, we dived deep into abstract algebra to motivate why Geometric Deep Learning is an interesting topic. Now we begin the journey to show that it is also useful in practice. In summary, we know that symmetries constrain our hypothesis class, making learning simpler—indeed, they can make learning a tractable problem. How does this happen?
Mathematical foundations for Geometric Deep Learning
Published:
Yes, abstract algebra is actually useful for machine learning.
LaTeX tricks
Published:
Improve typesetting and save space in your submissions, who does not want that?
2021
Bayesian Statistics - Techniques and Models flashcards
Published:
It’s again a statistics deck.
Pearls of Causality #11: Front- and Back-Door Adjustment
Published:
Two ways to shut the door before confounding enters the scene.
Pearls of Causality #10: Interventions and Identifiability
Published:
Interventions in disguise.
Pearls of Causality #9: Potential, Genuine, Temporal Causes and Spurious Association
Published:
Hitting the nail on its arrowhead, a.k.a. when does $X$ cause $Y$?
Pearls of Causality #8: Inferred Causation, $IC$, and ${IC}^*$
Published:
We will talk about IC, $IC$, and ${IC}^*$ in this post. You get the difference.
Pearls of Causality #7: Latent Structures and Stability
Published:
DAGs like to play hide-and-seek. But we are more clever.
Pearls of Causality #6: Markov Conditions
Published:
The model zoo of Markovian conditions is fascinating confusing. Let there be light!
Pearls of Causality #5: Statistical vs Causal Inference
Published:
What you won’t be able to find in this post are unconditional claims of superiority of causal inference.
Bayesian Statistics - From Concept to Data Analysis flashcards
Published:
This blog discusses causal inference. What is this post about Bayesian Statistics then?
Pearls of Causality #4: Causal Queries
Published:
Asking a causal question is not casual.
Pearls of Causality: The Causal Dictionary
Published:
No one told me that I need a dictionary for learning causal inference. Indeed, there was none before. Now there is.
Probabilistic Graphical Models 3 flashcards
Published:
Not just parameter learning, but learning about parameter learning got easier today.
Pearls of Causality #3: The properties of d-separation
Published:
A top-secret guide to d-separation. We will go deep, ready?
Pearls of Causality #2: Markov Factorization, Compatibility, and Equivalence
Published:
This post deliberately (wink) tries to confuse you about the grand scheme of DAG equivalence. What a good deal, isn’t it?
Probabilistic Graphical Models 2 flashcards
Published:
If your goal is to be able to recall Sum-Product Belief Propagation even at 3a.m., this is the post you are looking for.
Pearls of Causality #1: DAGs, d-separation, conditional independence
Published:
d-separation is the bread and butter for deciding about conditional independence in DAGs. What is a DAG, anyway?
A post series on Pearl: Causality
Published:
A causality blog cannot exist without discussing Judea Pearl’s Causality book. Thus, I am paying my debt.
Probabilistic Graphical Models 1 flashcards
Published:
To make learning probabilistic graphical models frictionless and more fun.
Causality resources
Published:
Good resources matter, a lot.
Welcome to my journey!
Published:
A PhD student’s casual journey with causal inference.