SGBAU · Third Year B.E. Artificial Intelligence and Data Science · NEP 2020
Machine Learning Techniques
5AD221PC · Semester V · 3 credits · IE 40 + ESE 60
Unit 1: Fundamentals of Neural Networks
8 hoursNeural networks and the human brain, artificial neuron models and the McCulloch–Pitts model, characteristics, learning methods and the taxonomy of architectures; the perceptron and its learning algorithm, linear separability and XOR, error correction and gradient descent, batch, stochastic and mini-batch learning, learning-rate schedules, momentum and Adam.
Examined in the 60-mark end-semester paper
- Biological and artificial neurons
- McCulloch–Pitts neuron and thresholding logic
- Perceptron and multilayer networks
- Gradient descent variants
- Momentum and Nesterov accelerated gradient descent
Unit 2: Backpropagation Networks
8 hoursSupervised learning with neural networks, the backpropagation architecture and the multilayer perceptron, the chain rule and the backpropagation algorithm, training and error calculation, steepest descent and the effect of the learning rate, loss functions and optimisation, and overfitting and generalisation.
Examined in the 60-mark end-semester paper
- Backpropagation in neural networks
- Loss functions in machine learning
- Bias-variance trade-off
Unit 3: Associative Memory and Adaptive Resonance Theory
6 hoursAuto- and hetero-correlation, multiple training encoding, the Hopfield network and its error performance, exponential BAM, associative memory for real-coded pattern pairs; adaptive resonance theory — cluster structure and category formation, ART1 and ART2 — and character recognition with ART1.
Examined in the 60-mark end-semester paper
- Hopfield networks and bidirectional associative memory
- Adaptive resonance theory: ART1 and ART2
Unit 4: Regression
6 hoursAn overview of supervised, unsupervised and reinforcement learning; linear regression with one variable, covariance and correlation, evaluating regression models and their metrics, gradient descent for linear regression, multiple and non-linear regression, time-series analysis and ARIMA, and hypothesis testing of regression and ARIMA models.
Examined in the 60-mark end-semester paper
- Types of learning
- Simple and multiple linear regression
- Polynomial regression
- Regression metrics
- ARIMA forecasting
Unit 5: Classification
6 hoursBinary classification with naive Bayes, k-nearest neighbours, decision trees and logistic regression; SVMs, ensembles, bagging and random forests; cross-validation and regularisation, performance measures, error analysis, the confusion matrix and ROC.
Examined in the 60-mark end-semester paper
- Bayesian learning and naive Bayes
- K-nearest neighbours
- Decision tree learning and overfitting
- Logistic regression
- Support vector machines
- Bagging and random forests
- Cross-validation and model evaluation
- Classification metrics
Unit 6: Clustering and Dimensionality Reduction
6 hoursClustering and distance measures, distance-, density-based and hierarchical methods, k-means with continuous and categorical values, Gaussian mixture models and the EM algorithm; eigen-analysis — SVD, PCA and LDA — and dimensionality reduction with applications in ML and NLP.
Examined in the 60-mark end-semester paper
- K-means and choosing k
- DBSCAN
- Hierarchical clustering
- Gaussian mixture models
- Principal component analysis
- Linear discriminant analysis
Unit contents are summarised from the official SGBAU syllabus . AiCanCode.org is not affiliated with Sant Gadge Baba Amravati University (SGBAU).