Teaching

10-701

○ Course taught by Dr. Zoltan Mikols.
○ The objective is to get acquainted with Deep Learning library - Tensorflow and Keras. In the class, the students will explore fully connected neural networks, CNN, RNN and LSTM variants. The idea is to build a smaller network first and then do the incremental change in the model. The students are required to observe the change in the model accuracy by varying optimisers (vanilla sgd (learning rates), adagrad, momemtum, adam), Loss functions, different regularisers (L1, L2, drop-out, Elastic net) to tackle overfitting, word embedding techniques, etc. The problems are based on image Classification, text preprocessing and sentiment analysis.
○ Dataset: CIFAR 10, IMDB, Ohsumed.

10-707

○ Course taught by Prof. Zoltan Mikols.
○ The objective is to learn the machine learning pipeline for data mining.The practical work will be given in Python and will explore standard machine learning techniques (Exploratory Data Analysis, Missing Values, Build Model, Cross-validation, Tune Parameters) that are widely used in academia and industry.
○ Dataset: Students will explore Titanic [Thanks to Kaggle] and Census Dataset [Thanks to UCI Machine learning repository and several contributors], build a model and tune the parameters to improve predictions.

○ Course taught by Prof. Raju Halder.
○ Course taught to under graduate students
○ The objective of the course was to have the basic understanding of shell programming and C language.