Back

Pim Ketras

Pim Ketras is a machine learning experiment that generates text based on a corpus of lyrics from Kim Petras's discography. It is Long Short-Term Memory (LSTM) neural network that was trained using TensorFlow.js. The website was built using React.

I built this to learn about machine learning while at Recurse Center.

I started by collecting lyrics from Kim Petras's discography.

Then, I trained the neural network using TensorFlow.js. Training an LSTM network involves teaching a model how to make accurate predictions based on sequential data. In this case, which words to predict based on patterns in the lyrics.

The model is trained by feeding it a sequence of words and then asking it to predict the next word. The model is then updated based on how accurate its prediction was. This process is repeated until the model is able to make accurate predictions.