NER Hyperparameter Tuning for Finance
Hyperparameter tuning is choosing a set of optimal hyperparameters for a learning algorithm. A hyperparameter is a model argument whose value is set before the learning process begins. The key to machine learning algorithms is to chose the optimal pair/pairs of hyperparameters.
For our project we just focus on tuning two parameters of the model:
1. Learning rate
2. Dropout
Created on May 11|Last edited on May 11
Comment
Highlights
- We use Roberta-base model for our NER tasks.
- We use the Spacy library for configuring and training the model with different parameters.
- We use Weights and Biases for logging the hyperparameter process.
- We evaluate the model performance using Precision, Recall and F1 Score.
- We only focus on models with high skill precision.
Tuning Parameters
- We use Bayesian Optimization techniques for tuning.
- We focus on tuning the learning_rate and dropout of the model.
Tuning results
Below we only focus on the charts which have the best skill precision and train the final models with these parameters to finally choose the best one.
Run set
3
Add a comment