**Course plan of BM616:** Applied Machine Learning

**Total lectures:** 28.

**Total labs:** 12

**Total credits:** 3

**Maximum student intake:** 50

**Priority of student enrollment:** BME M.Tech., PhD any branch (except CSE), M.Tech. any branch (Except CSE), B. Tech. any branch except CSE, Any program of CSE.

**Lab format:** Lab tasks shall be given beforehand as assignments and 1 hour of weekly contact in lab shall be provided. Weightage (30)

**Project:** Weightage (15 marks). Starts after mid-sem. One problem from kaggle (or similar data science platforms) needs to be solved. One Paper needs to written and one presentation to be given.

**Mid-sem: **Weightage (22.5)

**End-Sem:** Weightage (32.5)

Students are recommended to have their own laptops/PCs for assignments. GPU access can be given remotely from medical devices lab.

**Pre-requisites: **Basic familiarity with linear algebra and probability theory. Basic prior exposure to programming in any language will be assumed.

**Recommended books:**

- Machine Learning by Tom Mitchell
- Hands–On Machine Learning with Scikit–Learn and TensorFlow by Aurelien Geron.
- Generative Deep Learning: Teaching Machines to Paint, Write, Compose, and Play by David Foster

**List of Lectures:**

**Lab 1: **Installing Anaconda and running ipython notebook on Google colab. (By TAs)

**Lecture 2: **Linear algebra overview.

**Lecture 3: **implementation of linear algebra operations on Python

**Lab 2: **Hands-on practice of matrix handling through basic image processing operations on Python.

**Lecture 3:** Probability overview.

**Lecture 4:** Python overview and setting up programming environment.

**Lecture 5:** Python coding and how to apply most frequent programming and data operations.

**Lab 1: **Assignment related to data file reading and manipulation. Data frames. Image reading and matrix operations

**Lecture 6: **Linear regression

**Lecture 7: **Logistic regression

**Lab 2: **Implementation of linear regression on house price prediction.

**Lecture 8: **Bias variance tradeoff and regularization.

**Lecture 9: **Training, testing and validation sets, Confusion matrix, F-score, Choosing adequate number of training sets, Early stopping.

**Lab 3: **Logistic regression implementation on cancer diagnostics problem.

**Lecture 10:** Neurons and Neural Network introduction.

**Lecture 11:** Backpropagation in neural networks.

**Lab 4:** Hand-writing recognition from MNIST dataset. Building the neural network from first principle

**Lecture 12: **Building models using tensorflow and keras.

**Lecture 13: **DIfferent types of activation functions, softmax layer, batch normalization, dropout regularization.

**Lab 5:** Implementation of mnist handwriting recognition and cancer diagnostics using tensorflow.

**Lecture 14: **Unsupervised learning, K-Means, PCA.

**Lecture 15: **Support vector machines.

**Lab 6: **Implementation of K-Means algorithm for an unsupervised problem

**Lecture 16:** Convolutional neural networks, convolution, strides, maxpool, dense layers, activations.

**Lecture 18: **Optimization algorithms for deep neural networks.

**Lab 7: **Classification of images of cats and dogs using CNNs implemented in tensorflow.

**Lecture 19:** Hyperparameter tuning and batch normalization. Popular deep learning architectures.

**Lecture 20:** Performance evaluation and metrics for evaluating CNNs.

**Lab 8: **Transfer learning implementation on VGG-16 pretrained model.

**Lecture 21: **Using pre-trained models, Transfer learning, One shot learning.

**Lecture 22:** Regression in neural networks, Object localization and segmentation using RCNN and YOLO.

**Lab 9: **Implementation of Yolo for live detection and bounding box predictions of objects from webcam feed. Generation of a random circle and its detection.

**Lecture 23: **Recurrent neural networks (RNNS) and LSTMS and their applications.

**Lecture 24:** Time series and next word prediction using RNNS and LSTMS

**Lab 10: **Time series prediction of ECG samples, stock market data and music.

**Lecture 25:** Variational autoencoders

**Lecture 26:** Generative adversarial networks (GANs)

**Lab 11: **Encoding MNIST handwriting samples and CIFAR images followed by reconstruction and new image generation.

**Lecture 27:** Applications of GANs

**Lecture 28: **Conclusion and advanced applications of machine learning under research.

**Lab 12: **GAN based human speech generation.

**Recorded lectures:**