PAHSU · Third Year B.Tech. Computer Science and Engineering · NEP 2020
Artificial Neural Network
CSEPEC-03C · Programme Elective III · Semester VI · 4 credits · ISE 30 + ESE 70
One of the Programme Elective III options. Programme Elective III carries 3 lecture credits and 1 tutorial credit.
Unit 1: Introduction to ANN
7 hoursHistory of neural networks; structure and working of the biological neural network; neural net architectures and topologies; features, characteristics and types; activation functions; McCulloch-Pitts, perceptron and Adaline models; basic learning laws; applications; BNN versus ANN.
Examined in the 70-mark end-semester paper
- Artificial neuron models
- Neural network architectures
- McCulloch-Pitts, perceptron and Adaline models
- Basic learning laws
- Activation functions in deep networks
Unit 2: Learning Algorithms
8 hoursLearning and memory; the number of hidden nodes; error-correction and gradient-descent rules; perceptron learning; supervised backpropagation in multilayer networks; feed-forward and feedback networks.
Examined in the 70-mark end-semester paper
- Perceptron learning rule
- Linear separability and multilayer perceptrons
- Backpropagation training
Unit 3: Associative Learning
8 hoursAssociative learning; the Hopfield network and its error performance; simulated annealing; the Boltzmann machine and Boltzmann learning; state-transition diagrams and false minima; stochastic update; pattern association, classification and mapping.
Examined in the 70-mark end-semester paper
- Associative memory
- Bidirectional associative memory and Hopfield networks
- Hill climbing and simulated annealing
- Boltzmann machines and deep belief networks
- Pattern association, classification and mapping
Unit 4: Competitive Learning Neural Networks
8 hoursComponents of a competitive-learning network; pattern clustering and feature-mapping networks; ART networks and character recognition with ART; self-organising maps and the SOM algorithm; learning vector quantisation; adaptive pattern classification.
Examined in the 70-mark end-semester paper
- Competitive learning and ART networks
- Self-organising maps
- Learning vector quantisation
Unit 5: Convolutional Neural Networks
8 hoursBuilding blocks and architectures of CNNs: convolution and pooling layers, padding, strided convolutions and convolutions over volumes; softmax regression; deep learning frameworks; mismatched train and test distributions; transfer, multi-task and end-to-end learning; LeNet-5, AlexNet, VGG-16 and ResNet.
Examined in the 70-mark end-semester paper
- CNN architecture
- Convolution, padding and strides
- Pooling and fully connected layers
- Transfer learning and data augmentation
- LeNet and AlexNet
- CNN architectures: VGG, GoogLeNet, ResNet and DenseNet
Unit 6: Applications of ANN
6 hoursPattern classification: recognising Olympic-games symbols and printed characters; the neocognitron for handwritten characters; NETtalk for text to speech; recognising consonant-vowel segments; texture classification and segmentation.
Examined in the 70-mark end-semester paper
- Cognitron and neocognitron
- Neural network applications in pattern and speech recognition
Unit contents are summarised from the official PAHSU syllabus . AiCanCode.org is not affiliated with Punyashlok Ahilyadevi Holkar Solapur University (PAHSU).