1. scikit-learn CLICK
2. NumPy CLICK
3. SciPy CLICK
4. matplotlib
5. pandas CLICK
Supervised Learning Models -> CLICK
Unsupervised Learning Models -> CLICK
Semi-Supervised Learning-> CLICK
Reinforcement Learning -> CLICK
Deep Learning (Neural Networks) ->CLICK
Different type of models used in Machine Learning in Python -> CLICK
which model and when? -
Training data is processed using three primary machine learning frameworks:
1. supervised learning (using labeled data to predict outcomes),
2.unsupervised learning (finding hidden patterns in unlabeled data), and
3. reinforcement learning (learning via trial, error, and rewards).
START
│
Do you have Y?
/ \
YES NO
│ │
What type is Y? What do you need?
/ \ / | \
Category Number Groups Reduce Outliers
│ │ │ dimensions │
↓ ↓ ↓ ↓ ↓
Classification Regression Clustering PCA Isolation
│ │ │ Forest
↓ ↓ ↓
┌──────────┐ ┌───────────┐ ┌──────────┐
│ Logistic │ │ Linear │ │ K-Means │
│ Random │ │ Ridge │ │ DBSCAN │
│ Forest │ │ Random │ │ HDBSCAN │
│ SVM │ │ Forest │ │ GMM │
│ KNN │ │ Gradient │ │ Spectral │
│ Naive │ │ Boosting │ │ │
│ Bayes │ │ SVR │ │ │
└──────────┘ └───────────┘ └──────────┘
If your goal is to forecast a continuous, numerical value, you will use Supervised Regression models.
- Example Use Case: Predicting home prices, forecasting stock trends, or estimating delivery times.
- Common Models: Linear Regression, Random Forest Regressor, or Gradient Boosting (XGBoost).
2. Sorting Categories (Classification)
If your goal is to assign data points into distinct, predefined buckets, you will use Supervised Classification models.
- Example Use Case: Sorting emails into "Spam" or "Inbox", identifying if an image shows a cat or a dog, or predicting if a transaction is fraudulent.
- Common Models: Logistic Regression, Support Vector Machines (SVM), Random Forest Classifier, or Neural Networks.
3. Grouping Data (Clustering)
If your goal is to find hidden patterns and group data together without knowing the categories in advance, you will use Unsupervised Clustering models.
- Example Use Case: Segmenting customers by purchasing behavior, grouping similar news articles together, or finding anomalies in network traffic.
- Common Models: K-Means Clustering, Hierarchical Clustering, or DBSCAN.
| Feature / Capability | Scikit-learn | PyTorch | TensorFlow |
|---|---|---|---|
| Primary Focus | Traditional ML & Tabular Data | Deep Learning & Research | Production Deep Learning |
| Hardware | CPU Only | GPU & TPU Acceleration | GPU & TPU Acceleration |
| Data Scaling | Small to Medium (Spreadsheets/CSVs) | Massive (Images, Audio, Text) | Massive (Big Data pipelines) |
| Graph Type | Static Execution (Pre-built APIs) | Dynamic Graph (Pythonic/Flexible) | Static Graph (Highly Optimized) |
| Code Complexity | Low (model.fit() & model.predict()) | Medium to High (Custom training loops) | Medium (via high-level Keras API) |
| Model Family | Interpretability | Handles High Dimensions? | Training Speed | Primary Weakness |
|---|---|---|---|---|
| Linear Models (e.g., Logistic Regression) | High (Direct feature weights) | No (Prone to underfitting) | Fast | Cannot capture complex nonlinear rules |
| Tree/Ensemble Models (e.g., Random Forest) | Medium (Feature importance scores) | Moderate | Medium | Large memory footprint for deep trees |
| Distance-Based Models (e.g., K-Nearest Neighbors) | Low (Instance-dependent) | No (Suffers from "curse of dimensionality") | Instant (No active training phase) | Inference slows down heavily as dataset grows |
| Kernel/Matrix Methods (e.g., Support Vector Machines) | Low (Black-box decision boundaries) | Yes | Slow on large data | Computationally expensive tuning (O(n³) scaling) |
| Neural Networks (Deep Learning) | None (Millions of abstract weights) | Yes | Very Slow (Requires epochs on GPUs) | Prone to overfitting without massive training data |
Models Uset to train:
1. Classification (Sorting Categories)
Classification predicts categorical outputs, meaning it assigns data into predefined classes like spam/non-spam emails or disease risk categories. These algorithms learn to map input features to discrete labels. Here are some classification algorithms:
Logistic Regression
Decision Tree
Random Forest
K-Nearest Neighbors (KNN)
Naive Bayes
Support Vector Machine
2. Regression(Predicting Numbers)
Regression, predicts continuous values, such as house prices or product sales. It learns the relationship between input features and a numerical target variable. Here are some regression algorithms:
Linear Regression
Polynomial Regression
Ridge Regression
Lasso Regression
Decision tree
Random Forest
1. Clustering (Grouping Unlabeled Data)
Clustering is the process of grouping data points into clusters based on their similarity. This technique is useful for identifying patterns and relationships in data without the need for labeled examples. Common techniques include:
K-Means
DBSCAN
Mean-shift
2. Dimensionality Reduction Techniques
Dimensionality reduction helps reduce the number of features while preserving important information. Common techniques include:
Principal Component Analysis
Independent Component Analysis
3. Association Rule Learning
Association rule learning is a technique for discovering relationships between items in a dataset. It identifies rules that indicate the presence of one item implies the presence of another item with a specific probability. Common techniques include:
Apriori
FP-growth
Eclat
Here are some of most common reinforcement learning algorithms:
Q-learning: Learns the best action for each state based on expected rewards.
SARSA (State-Action-Reward-State-Action): Similar to Q-learning but updates values for the action actually taken.
Deep Q-learning: Uses neural networks to handle complex state-action relationships
Types of Reinforcement Learning
Positive Reinforcement: Rewards desired behavior (e.g., giving points for correct answers).
Negative Reinforcement: Removes negative outcomes to encourage good actions (e.g., turning off a buzzer after the right move).
No comments:
Post a Comment