APPM 5720 Reports

This repository contains biweekly reports for the Applied Deep Learning (APPM 5720) class taught by Dr. Maziar Raissi.

The contents of the reports are as follows:

Report 1

• Setting up my developer environment with Tensorflow and following Tensorflow tutorials.

Report 2

• Finish up Tensorflow tutorials and do exploratory dataset analysis on Imagenet.

Report 3

• I explore data augmentation, transfer learning, and overfitting.

Report 4

• I explore transfer learning in PyTorch, more data augmentation, and a combination of the two methods.

Report 5

• I explore knowledge distillation on both a Multi-Layer Perceptron and on a convolutional neural network.

Report 6

• I use two different Bayesian Optimization packages in different settings (hyperparameter tuning as well as function minimization).

Report 7

• I perform exploratory data analysis on the Cityscapes dataset.

Report 8

• I explore the SemEval-2012 Task 2 dataset used for NLP models in evaluating semantic relationships.

Report 9

• I explore the vectors from the GloVe model by utilizing cosine similarity as well as t-SNE.

Report 10

• I do EDA on the CoNLL-2003 dataset and use a Named Entity Recognition (NER) model.

Report 11

• For this report I explore Byte Pair Encoding (BPE) for tokenization.

Report 12

• For this report I explore and use BERT through the RoBERTa model.

Report 13

• For this report, I explore the CLIP model through downstream applications.

Report 14

• For this report, I analyze the effects of generation strategies on GPT2.