Deep Learning

This repository is used for the assignments' solution of "Introduction to Deep Learning" course offered at Otto-von-Guericke-Universität Magdeburg, Germany.

---

Short description of each task

Details are available inside each Assignment folder

Implemented in Tensorflow 2.0 and Python 3.

---

Assignment 1

---

For MNIST dataset, implemented at a very low level (custom training loop with tf.GradientTape()) using Tensorflow 2.0 without keras functionality, initializing own weights and biases.

Assignment 2

---

1. Tensorboard integration
2. Solving erros in 5 files with explanation
3. Classification of MNIST using tf.data module and low level implementation
4. Playing around with shuffle, repeat and batch

Assignment 3

---

CNN built with certain keras functionality trained on MNIST, CIFAR10, CIFAR100 but using custom training loop with tf.GradientTape()

Assignment 4

---

1. CNN for classification using tf.function decorator to speed up process for training and test phase.
2. DenseNet implementation from scratch
3. High-level Training Loops with Keras (CNN)
4. TensorBoard Computation Graphs