sergiopperez

BBC_News_NLP

Classification model for the BBC news archive. Analysis of embedding layer to visualise word impact in classification

Cahn_Hilliard_Finite_Volume

Finite-volume schemes for the Cahn-Hilliard equation that unconditionally and discretely satisfy the boundedness of the phase field and the free-energy dissipation. Our scheme is applicable to a variety of free-energy potentials including the Ginzburg-Landau and Flory-Huggins, general wetting boundary conditions accounting for wall free energies, and degenerate mobilities.

Django_ML_Web_Deployment

ML-based web application using Django. Classification model for Titanic dataset and web application to accept inputs from user and return prediction

hackathon_podcast

Transcribir podcasts en español y aplicar PLN para resumir y responder preguntas

Mean_Housing_Prices

Run a t-test to show that university towns have their mean housing prices less affected by recessions. Data preprocessing with pandas, statistical analysis, t-test, hypothesis testing

Predict_Fraud

Train several models and evaluate how effectively they predict instances of fraud in bank transactions. Support vector classification, logisitic regression classifier, grid-search optimization

Property_Maintenance_Fines

To predict whether a given building maintenance fine will be paid on time. Models: random forest, gradient boosted decision trees, linear regression optimized with regularization and c parameter

Residual_Network

Create a residual network architecture with the model subclassing API from Tensorflow. Application to MNIST dataset

Reuters_Dataset_Tensorflow

Model subclassing and custom training loops in Tensorflow for the Reuters dataset. Natural language processing and document classification

Time_Series_Forecast_Neural_Networks

Bidirectional LSTM model to forecast a synthetic time series. Adaptive learning rate and Lambda layers to improve training stage

sergiopperez/Cahn_Hilliard_Finite_Volume

Finite-volume schemes for the Cahn-Hilliard equation that unconditionally and discretely satisfy the boundedness of the phase field and the free-energy dissipation. Our scheme is applicable to a variety of free-energy potentials including the Ginzburg-Landau and Flory-Huggins, general wetting boundary conditions accounting for wall free energies, and degenerate mobilities.

sergiopperez/hackathon_podcast

Transcribir podcasts en español y aplicar PLN para resumir y responder preguntas

sergiopperez/Time_Series_Forecast_Neural_Networks

Bidirectional LSTM model to forecast a synthetic time series. Adaptive learning rate and Lambda layers to improve training stage

sergiopperez/Reuters_Dataset_Tensorflow

Model subclassing and custom training loops in Tensorflow for the Reuters dataset. Natural language processing and document classification

sergiopperez/BBC_News_NLP

Classification model for the BBC news archive. Analysis of embedding layer to visualise word impact in classification

sergiopperez/Django_ML_Web_Deployment

ML-based web application using Django. Classification model for Titanic dataset and web application to accept inputs from user and return prediction

sergiopperez/Image_Inpainting

sergiopperez/Residual_Network

Create a residual network architecture with the model subclassing API from Tensorflow. Application to MNIST dataset

sergiopperez/Mean_Housing_Prices

Run a t-test to show that university towns have their mean housing prices less affected by recessions. Data preprocessing with pandas, statistical analysis, t-test, hypothesis testing

sergiopperez/Predict_Fraud

Train several models and evaluate how effectively they predict instances of fraud in bank transactions. Support vector classification, logisitic regression classifier, grid-search optimization

sergiopperez/Property_Maintenance_Fines

To predict whether a given building maintenance fine will be paid on time. Models: random forest, gradient boosted decision trees, linear regression optimized with regularization and c parameter

sergiopperez/Automatic_Differentiation_Tensorflow

Automatic differentiation for linear regression in Tensorflow. Model subclassing, `tf.GradientTape()`, training linear regression model

sergiopperez/Cohere_API_projects

Experimenting with Cohere's platform to build NLP applications

sergiopperez/Flask_Tweet_Classification_Deployment

sergiopperez/GNNs_experiments

A bunch of notebooks taken from PyTorch Geometric: https://pytorch-geometric.readthedocs.io/en/latest/notes/colabs.html

sergiopperez/hyde

A brazen two-column theme for Jekyll.

sergiopperez/lm-evaluation-harness

A framework for few-shot evaluation of language models.

sergiopperez/python-cam

Program for calculating contact angles from an image or video of a sessile drop.

sergiopperez/sergiopperez.github.io

sergiopperez/Text_Prediction_Shakespeare

Text prediction model from Shakespeare's sonnets. Tokenize, pad, model with embedding and stacked RNN, validation and prediction.

sergiopperez/Time_Series_Forecast_Statistics

ARMA model to forecast time series. Dickey-Fuller test for stationarity and differencing to remove trend.

sergiopperez/transformers

🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.

sergiopperez/Well_Balanced_Finite_Volume

Well-balanced finite volume scheme for hydrodynamic equations with a general free energy. Data visualization, scientific computing, temporal evolution of mathematical models from fluid dynamics