Algorithms & Tasks

A collection of notebooks covering fundamental machine learning algorithms and image processing techniques — built as a learning portfolio.

---

What's Inside

Machine Learning

Notebook	Topic	Key Concepts
01_linear_regression.ipynb	Linear Regression	From-scratch implementation, gradient descent, cost function
02_classification_algorithms.ipynb	Classification	Logistic Regression, KNN, SVM, Naive Bayes
03_trees_and_ensembles.ipynb	Tree-based Models	Decision Trees, Random Forest, Gradient Boosting
04_clustering.ipynb	Unsupervised Learning	K-Means, Hierarchical Clustering, DBSCAN
05_pca_dimensionality_reduction.ipynb	Dimensionality Reduction	PCA, t-SNE, explained variance

Image Processing

Notebook	Topic	Key Concepts
06_image_basics.ipynb	Image Fundamentals	Pixel arrays, color spaces, transformations
07_filters_edge_detection.ipynb	Filtering	Convolution, Gaussian blur, Sobel, edge detection
08_image_segmentation.ipynb	Segmentation	Thresholding, K-Means segmentation, morphological ops

Java Algorithms

File	Algorithm
MergeSort.java	Merge Sort implementation
QuickSort.java	Quick Sort implementation
FrostBite.java	OOP practice (Codemon character class)

---

What Each Notebook Covers

Every notebook follows the same structure:

1. Introduction — what the algorithm does and why it matters
2. Data Loading — real datasets from sklearn
3. EDA — visualizations to understand the data
4. Implementation — the algorithm (sometimes from scratch)
5. Evaluation — metrics, confusion matrices, comparisons
6. What I Learned — key takeaways

---

Installation

git clone https://github.com/naveed1011/algorithms_and_tasks.git
cd algorithms_and_tasks
pip install numpy pandas matplotlib seaborn scikit-learn scipy jupyter
jupyter notebook

---

Key Takeaways

From ML Notebooks

• No free lunch theorem — no single algorithm works best for all problems
• Feature scaling matters — especially for distance-based algorithms (KNN, SVM)
• Always cross-validate — single train/test splits can be misleading
• Ensemble methods usually win — Random Forest and Gradient Boosting are strong baselines
• Interpretability vs performance — decision trees are interpretable, ensemble methods perform better

From Image Processing Notebooks

• Images are just arrays — once you see them as numbers, everything makes sense
• Filters are convolution — sliding a small matrix over the image
• Noise reduction before edge detection — always smooth first
• Segmentation has many approaches — threshold, clustering, region growing
• Morphological operations are useful — clean up binary masks

---

Datasets Used

Dataset	Source	Task
California Housing	sklearn	Regression
Breast Cancer	sklearn	Binary Classification
Wine	sklearn	Multi-class Classification
Iris	sklearn	Clustering
Digits	sklearn	Dimensionality Reduction

---

Author

Naveed Ahmad

Learning ML and Image Processing, one notebook at a time.