Scribd
Upload
9 views2 pages

Machine Learning Fundamentals Explained

The document provides an overview of machine learning fundamentals, distinguishing between supervised and unsupervised learning, along with common algorithms for each. It discusses model evaluation metrics, the importance of training, validation, and test sets, and the concepts of overfitting and underfitting. Additionally, it introduces deep learning concepts, including neural networks and various architectures like CNNs and RNNs.

Uploaded by

letanphat21062004
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views2 pages

Machine Learning Fundamentals Explained

The document provides an overview of machine learning fundamentals, distinguishing between supervised and unsupervised learning, along with common algorithms for each. It discusses model evaluation metrics, the importance of training, validation, and test sets, and the concepts of overfitting and underfitting. Additionally, it introduces deep learning concepts, including neural networks and various architectures like CNNs and RNNs.

Uploaded by

letanphat21062004
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Machine Learning Basics

1. Supervised vs. Unsupervised Learning

Supervised Learning:
- Uses labeled data (input and correct output)
- Goal: predict future outcomes or classify into categories
- Common Tasks: Classification (e.g., spam detection), Regression (e.g., house price prediction)

Unsupervised Learning:
- Uses only input data without labels
- Goal: discover hidden patterns or structures in data
- Common Tasks: Clustering (e.g., customer segmentation), Dimensionality Reduction (e.g., PCA)

2. Common Algorithms (High Level)

Supervised Algorithms:
- Linear Regression: Predicts continuous values.
- Logistic Regression: Binary/multiclass classification.
- Decision Trees / Random Forests: Tree-based models; handle non-linearities well.
- SVM: Finds optimal separating hyperplane.
- KNN: Predicts by finding the nearest neighbors.

Unsupervised Algorithms:
- K-Means Clustering: Groups data into k clusters.
- Hierarchical Clustering: Builds nested clusters.
- PCA: Reduces feature space while preserving data variance.

3. Model Evaluation Metrics

Classification Metrics:
- Accuracy: (TP + TN) / Total samples
- Precision: TP / (TP + FP)
- Recall: TP / (TP + FN)
- F1 Score: Harmonic mean of Precision and Recall
- Confusion Matrix: Tabular layout of TP, TN, FP, FN

Regression Metrics:
- MSE (Mean Squared Error)
- RMSE (Root MSE)
Machine Learning Basics

- MAE (Mean Absolute Error)


- R² Score: Goodness of fit measure

4. Training, Validation, and Test Sets

- Training Set: For learning model parameters (60-80%)


- Validation Set: For tuning hyperparameters (10-20%)
- Test Set: For evaluating model's generalization (10-20%)

Typical split: 70% Train / 15% Validation / 15% Test

5. Overfitting vs. Underfitting

Overfitting:
- Model memorizes training data and performs poorly on unseen data.
- Symptoms: High training accuracy, low test accuracy.
- Solutions: Simpler model, regularization (L1/L2), more data, dropout.

Underfitting:
- Model is too simple to learn patterns in data.
- Symptoms: Low accuracy on both train and test sets.
- Solutions: More complex model, better feature engineering.

6. Deep Learning Overview

- Neural Networks: Layers of neurons to learn representations.


- Activations: ReLU, Sigmoid, Tanh introduce non-linearity.
- Optimizers: SGD, Adam adjust weights using gradients.
- CNNs: Great for images; use filters to capture spatial features.
- RNNs / LSTMs / GRUs: Ideal for sequences (text, time series).
- Techniques: Dropout, Batch Normalization, Data Augmentation.
- Transfer Learning: Use pre-trained models to boost performance with limited data.

You might also like