AI Related

🧮 From Scratch

Every algorithm implemented in pure NumPy — no scikit-learn, no PyTorch. Understand the math before the framework.

🛍️ Real Data

Microsoft Surface product data as the consistent applied context across all 15 labs.

✅ Verified

Every code block runs in Docker. Every output block shows real verified results.

Prerequisites: Python Foundations + Practitioner. All labs use only numpy 2.4.2 and pandas 3.0.1 — no scikit-learn or PyTorch required.

Quick Start

🐳 Docker (Recommended)

docker pull zchencow/innozverse-python:latest
docker run --rm zchencow/innozverse-python:latest python3 -c "import numpy; print(numpy.__version__)"

Local

---

Lab Curriculum

#	Lab	Topics	Time
01	Linear Regression	MSE, gradient descent, R², normalisation	35 min
02	Logistic Regression	Sigmoid, binary cross-entropy, L2, confusion matrix	30 min
03	Neural Network from Scratch	Backprop, ReLU, softmax, He init	40 min
04	K-Means Clustering	K-Means++, elbow, silhouette, market segmentation	30 min
05	Decision Trees	Gini impurity, information gain, pruning, feature importance	35 min
06	K-Nearest Neighbours	Distance metrics, weighted voting, ANN context	25 min
07	PCA	Covariance, eigenvectors, explained variance, reconstruction	30 min
08	NLP & TF-IDF	Tokenisation, TF-IDF, cosine similarity, Naive Bayes	30 min
09	Recommendation Systems	Content-based, collaborative filtering, SVD, hybrid	30 min
10	Time Series	SMA/EMA, decomposition, ACF, AR(p) forecast	30 min
11	Gradient Descent Optimisers	SGD, Momentum, RMSProp, Adam, bias correction	30 min
12	Convolutional Operations	Conv2D, Sobel, Gaussian, pooling, stride/padding	35 min
13	Data Preprocessing	Imputation, IQR, scaling, encoding, feature selection	35 min
14	Model Evaluation	K-fold CV, stratified CV, metrics, bias-variance	35 min
15	Capstone: Full ML Pipeline	End-to-end: ingest → preprocess → train → ensemble → serialise	45 min

Total: ~500 minutes of hands-on AI/ML from scratch