KBCNMU · Third Year B.E. Computer Engineering · NEP 2020
Deep Learning and Neural Network
CS-606 B · Elective III · Semester VI · 3 credits · ISE 40 + ESE 60
One of the Elective III options.
Unit 1: Neural Networks and Rosenblatt's Perceptron
9 hoursWhat a neural network is, the human brain, models of a neuron, networks as directed graphs, feedback and architectures; the perceptron, its convergence theorem, and its relation to the Bayes classifier.
Examined in the 60-mark end-semester paper
- Artificial neuron models
- Neural network architectures
- Perceptron learning rule
- McCulloch-Pitts, perceptron and Adaline models
Unit 2: Supervised and Unsupervised Neural Networks
9 hoursArchitecture and training of Hebb, perceptron, back-propagation, associative-memory, Hopfield, counter-propagation and adaptive-resonance networks.
Examined in the 60-mark end-semester paper
- Basic learning laws
- Backpropagation training
- Associative memory
- Bidirectional associative memory and Hopfield networks
- Counterpropagation networks
- Competitive learning and ART networks
Unit 3: Advanced Neural Networks
9 hoursKohonen self-organising maps, learning vector quantisation, the Boltzmann machine, cognitron and neocognitron, and optical neural networks with electro-optical multipliers and holographic correlators.
Examined in the 60-mark end-semester paper
- Self-organising maps
- Learning vector quantisation
- Boltzmann machines and deep belief networks
- Cognitron and neocognitron
Unit 4: Deep Learning
9 hoursMachine-learning basics and linear regression, under- and over-fitting, support vector machines, deep feed-forward, convolutional and recurrent networks, deep Boltzmann machines, and applications in speech recognition and NLP.
Examined in the 60-mark end-semester paper
- Deep learning vs machine learning
- Support vector machines and kernels
- Perceptron and feed-forward networks
- CNN architecture
- Recurrent neural networks
Unit 5: GANs, Training and Deployment
9 hoursEncoder/decoder and generator/discriminator architectures; challenges in training — data augmentation and hyperparameters; transfer learning; developing and deploying models with TensorFlow or PyTorch.
Examined in the 60-mark end-semester paper
- Generative adversarial networks
- Encoder–decoder sequence-to-sequence models
- Transfer learning and data augmentation
- Hyperparameters: learning rate, momentum and regularisation
- MLOps and model deployment
Unit contents are summarised from the official KBCNMU syllabus . AiCanCode.org is not affiliated with Kavayitri Bahinabai Chaudhari North Maharashtra University (KBCNMU).