r/365DataScience

Dear Folks, I have created multiple content on Machine Learning(work in progress), and they are free. I am a data scientist and a post grad degree holder in AI/ML from IIT. To help the machine learning community with important Machine Learning Concepts, I have created multiple long form videos, and structured topicwise digestible contents structured as playlists for learning.

If you go through the first two playlists:

Introductory Machine Learning Concepts
Probability Foundations: Univariate Models

You might find helpful content, I have tried explaining with intuitions, derivations, and this is work in progress. For code implementations, scikit learn website has great content on them as well. In total they have 60+ topicwise videos so far, and I think they have the potential to help folks a lot in starting with concepts, or getting with mathematical concepts, or whether you are preparing for an AI/ML/Data job interviews etc.

When I sat for my interviews, I was grilled on my project, but majority of questions from my project tested more on foundational concepts and there know how’s.

These are FREE content on youtube. This is for the benefit of the learning community.

Link: https://youtube.com/@aayushsugandh4036?si=w5MKORU2fWzLRrAJ

Solve murders. Master SQL. One query at a time.

Agatha Christie cases. Real suspects. Live SQLite database. You write the queries, you catch the killer.

SELECT s.name, a.location
FROM suspects s
JOIN alibis a ON s.suspect_id = a.suspect_id
WHERE a.time_from <= '23:00'
AND a.location != 'Cabin'
ORDER BY s.name;

That's the kind of query standing between you and the murderer.

No signup. Runs in the browser. → querythemurder.com

Feedback: querythemurder@gmail.com

Hello Folks,

Have you ever wondered why we use sigmoid function so often in Machine Learning? Although it gives us a probability, it comes from Exponential families, and this exponential family, subsumes many of the distributions, that we study in Machine Learning.

In this lecture, we understand exponential families, Directional derivatives(Gradients and Hessians), study mixture Models, and understand how domain knowledge in Probabilistic Graphical Models makes our life simpler to model joint probability densities.

Timeline breakup(in hours and minutes):
0:00-0:17 - Understanding exponential families.
0:17-0:27 - Deriving Sigmoid Function for Bernoulli.
0:27-0:48 - Understanding log partition function, convex functions and proving why positive definite of hessians imply convexity, and why convex needed?
0:48-1:04 - Directional derivates(deriving gradients and hessians)
1:04-1:26 - Maximum entropy derivation of the exponential family.
1:26-1:56 - Mixture Models(Gaussians and Bernoulli Mixture Models)
1:56-2:16 - Probabilistic Graphical Models
2:16-2:34 - Markov Chains
2:34-End - Inference and Learning, Plate Notation diagram of Gaussian Mixture Models.

If you have watched earlier of my lectures from the playlist, they will help. I try explaining as if I am a learner, to simplify complex concepts. Everything I write in whiteboard, and these are completely FREE lectures to mention.

Link: https://youtu.be/T1uTBtJ7aHU?si=rozXSTjtSqPaaYb5