AM
Aarav Mehta
last week
Concise capsules I can finish over coffee, yet each one taught me something I actually use.
Free preview available. Sign in and subscribe to unlock the full book.
Vizuara AI Labs · beginner
Neural Networks from Scratch
Build a working neural network from an empty file.
From matrix multiplication to a trained classifier: softmax, cross-entropy, the forward and backward pass, and optimizers, all coded by hand in NumPy. No autograd, no framework magic — just the math, made visible and built up one layer at a time.
33 capsules118 figures~6 hoursby Dr. Raj Dandekar
Read on your Kindle or e-reader
Download the EPUB and read offline — perfect for the train. Works on Kindle, Kobo, Apple Books & more.
00The Ten Questions4 capsules
The Ten Questions — 4 chapters.
01Introduction and the Ten Questionsintuitionfree15 min02What a Neural Network Must Actually Doconceptfree11 min03A Mathematical Function to Switch a Neuron On or Offmath🔒13 min04Matrix Multiplication as a Classifiermath🔒11 min01Turning Scores into Probabilities3 capsules
Turning Scores into Probabilities — 3 chapters.
05From Raw Scores to a Probability Distributionintuition🔒11 min06The Softmax Functionmath🔒12 min07Tensors: The Data Containersconcept🔒11 min02Weights and Wiring4 capsules
Weights and Wiring — 4 chapters.
08The Logic for Calculating Weight Valuesmath🔒11 min09Setting Up the Neural Networkcode🔒11 min10Activation Functionsconcept🔒11 min11The Network as One Big Functionintuition🔒11 min03How a Network Learns6 capsules
How a Network Learns — 6 chapters.
12Teaching, or Training, a Neural Networkintuition🔒11 min13The Forward Pass and Cross-Entropy Lossmath🔒12 min14The Backward Pass: The Intuitionintuition🔒12 min15The Backward Pass: Gradients by Handmath🔒12 min16Weight Updation with Gradient Descentmath🔒13 min17Hyperparameter Tuningconcept🔒15 min04Context and Review3 capsules
Context and Review — 3 chapters.
18A Glimpse of Convolutional Neural Networksconcept🔒12 min19Relooking at the Ten Questionsintuition🔒11 min20Masterclass: Building the Whole Thing in Codeintuition🔒11 min05Coding the Layers7 capsules
Coding the Layers — 7 chapters.
21The Problem: A Spiral Datasetproject🔒11 min22Designing the Dense Layer Classcode🔒11 min23Coding the Layer Forward Passcode🔒11 min24Coding the Layer Backward Passcode🔒10 min25Adding Activation Functionscode🔒11 min26ReLU: Forward and Backward Passcode🔒12 min27Softmax: The Forward Pass in Codecode🔒11 min06Loss, Optimizer, and Training6 capsules
Loss, Optimizer, and Training — 6 chapters.
28Cross-Entropy Loss: The Forward Passcode🔒11 min29The Combined Softmax + Loss Backward Passmath🔒11 min30Optimizers: Beyond Plain Gradient Descentconcept🔒11 min31Coding the Optimizercode🔒9 min32Training the Entire Networkproject🔒11 min33Visualizing the Resultsproject🔒11 minRatings & reviews
1,891 readersRate this book
DO
Daniel Okafor
3 months ago
Rare mix of depth and readability. The worked examples are the clearest I have seen on this subject.
LC
Liam Chen
6 months ago
Best explanation of this topic I have found anywhere. The sidenotes fill in exactly the gaps I had.
MB
Marcus Bell
8 months ago
Dense but never confusing. Every chapter builds cleanly on the last — you can feel the care in the structure.