Machine Learning Basics

Machine Learning म्हणजे explicit rules न लिहता data मधून patterns शिकणे. Traditional programming मध्ये programmer rules लिहतो. ML मध्ये data आणि answers देतो, algorithm स्वतः rules शिकतो.

Traditional Programming:
Input + Rules → Output
"जर email मध्ये 'prize' किंवा 'free' असेल तर spam"
Problem: Rules manually लिहाव्या लागतात, limited

Machine Learning:
Input + Output → Rules (शिकतो)
1000 spam emails + 1000 valid emails दाखव
→ Model स्वतः patterns शिकतो
→ नवीन email spam आहे का ते सांगतो

1. Supervised Learning — Label सोबत शिकणे

प्रत्येक training example ला correct answer (label) आहे. Model शिकतो input → output mapping.

• ▸Classification — Output एका category मध्ये. Spam/Not Spam, Cat/Dog, Cancer/No Cancer.
• ▸Regression — Output एक number. House price predict करणे, temperature forecast.
• ▸Examples: Email spam filter, image recognition, loan approval

2. Unsupervised Learning — Labels नाहीत

Data मध्ये labels नाहीत — model स्वतः patterns आणि structure शोधतो.

• ▸Clustering — similar data points group करणे. Customer segmentation, news grouping.
• ▸Dimensionality Reduction — data simplify करणे. PCA.
• ▸Anomaly Detection — unusual patterns शोधणे. Fraud detection.

3. Reinforcement Learning — Trial & Error

Agent environment शी interact करतो, rewards/penalties मिळवतो, optimal behavior शिकतो. Chess AI, game playing (AlphaGo), robotics.

ML Workflow

1. Data Collection — relevant data गोळा करा
2. Data Preprocessing — clean, handle missing values, normalize
3. Feature Engineering — relevant features select/create करा
4. Model Selection — algorithm निवडा (Linear Regression, SVM, etc.)
5. Training — model data वर train करा
6. Evaluation — accuracy, precision, recall measure करा
7. Hyperparameter Tuning — model improve करा
8. Deployment — production मध्ये serve करा