Skip to Content
03 MathsDli Book

Practice Labs: Deep Learning Interviews (Kashani)

Source PDF: 2201.00650v2.pdf Book: Deep Learning Interviews by Shlomo Kashani (2nd Edition)

This book covers mathematical foundations and interview-style problems for deep learning. Labs below are organized by the book’s progressive difficulty levels.

Use this folder as a compact practice layer when you want to sharpen deep-learning intuition quickly. It works well as reinforcement for interviews, revision, or a short math-to-implementation bridge.

Labs

LabTopicBook ChapterDifficulty
Lab 01Logistic Regression from ScratchPart II: Logistic RegressionKindergarten
Lab 02Information Theory & EntropyPart III: Information TheoryHigh School
Lab 03Calculus, Gradients & BackpropagationPart III: Calculus, Algorithmic DifferentiationHigh School
Lab 04Probability & Bayesian Deep LearningPart II: Probabilistic Programming & Bayesian DLKindergarten
Lab 05Neural Network EnsemblesPart IV: NN EnsemblesBachelors
Lab 06CNN Feature Extraction & Deep LearningPart IV: CNN Feature Extraction + Deep LearningBachelors

How to Use

  1. Each lab is a Jupyter notebook with theory (markdown) and fully implemented code cells
  2. Read the theory cells, study the implementations, and run each cell
  3. Open in Jupyter: jupyter notebook lab_01_logistic_regression.ipynb

Prerequisites

  • Python 3.8+
  • NumPy
  • Matplotlib
  • SciPy (for Lab 04)

Suggested Order

  1. Lab 01 - Logistic Regression (foundational)
  2. Lab 04 - Probability & Bayesian DL (builds on probability basics)
  3. Lab 02 - Information Theory (entropy, KL divergence)
  4. Lab 03 - Calculus & Backpropagation (core for training NNs)
  5. Lab 06 - CNN Feature Extraction (applies NN concepts)
  6. Lab 05 - Ensemble Methods (advanced techniques)

How To Use This Folder Well

  • Treat these labs as short, focused reinforcement rather than a full standalone curriculum.
  • Use them when you want to turn abstract neural-network math into concrete coding exercises.
  • Revisit the relevant foundational or neural-network phases whenever a lab exposes a deeper gap.

What Comes Next

Last updated on