This is a repository of my learning about machine learning and other different topics associated to the broad field of AI.
- Capsule Network: CapsNet is a type of neural network that uses bunch of neurons to represent the internal structure of an image. It overcomes the shortcomings of the CNN. Paper
- Byte level encoding: Encoding is very important when it comes to LLMs. This GPT-2 paper introduces the concept of the byte level encoding algorithm. Paper
- Transformers: Paper
- Backpropagation: Learning representation by backpropagating errors Paper
- Batch normalization: Batch normalization is a method used to make training of artificial neural networks faster and more stable. It is a technique used to normalize the inputs of each layer in such a way that they have a mean output activation of zero and a standard deviation of one. Paper
- Covid 19 tweets classification: Text classification tasks related to tweets about the Coronavirus. The data has been gathered from Twitter, and manual tagging has been performed on the tweets. To protect privacy, the names and usernames have been replaced with codes. Link
- Natural disaster tweets classification: Prediction of which Tweets are about real disasters and which ones are not. Link
- Clean vs Dirty plates classification - Few shot learning: Train a classifier to differentiate between clean and dirty plates using limited labeled data. With only 20 clean and 20 dirty plates for training, perform a few-shot learning on hundreds of unlabeled plates to save time in dishwashing! Link
- Detecting contradiction and entailment in text - Natural language inferencing: In this project, the goal is to develop a Natural Language Inference (NLI) model capable of classifying pairs of premises and hypotheses into three categories: entailment, neutral, or contradiction. With text data available in fifteen different languages, the project offers an opportunity to explore and implement NLP techniques. Link