/MachineLearning

My machine learning portfolio

Primary LanguageJupyter NotebookMIT LicenseMIT

Machine learning Portfolio

This is a repository of the projects I worked on or currently working on. It is updated regularly. I also have the Jupyter Notebook version of some of my Kaggle kernels here.

Projects:

Kaggle kernels

CommonLit Readability Prize

  • CommonLit Readability - EDA & RoBERTa TF baseline [link]
  • Keywords(Text data, Deep learning, Transformers)

Cassava Leaf Disease Classification

  • Cassava Leaf Disease - TPU Tensorflow - Training [link]
  • Cassava Leaf Disease - TPU Tensorflow - Inference [link]
  • Cassava Leaf Disease - Training with TPU v2 Pods [link]
  • Cassava Leaf Disease - TPU v2 Pods - Inference [link]
  • Cassava Leaf - Supervised Contrastive Learning [link]
  • Cassava-Supervised Contrastive Learning-Inference [link]
  • Cassava Leaf Disease - CycleGAN data augmentation [link]
  • Cassava Leaf Disease-Stratified TFRecords 256x256 [link]
  • Keywords(Computer vision, Deep learning, Cassava, GPU, TPU, Supervised Contrastive Learning, GAN, CycleGAN, TFRecord)

Rainforest Connection Species Audio Detection

  • Rainforest-Audio classification Tensorflow starter [link]
  • Rainforest-Audio classification TF Improved [link]
  • Keywords(Audio data, Computer vision, Deep learning, Spectrogram)

I’m Something of a Painter Myself

  • Introduction to CycleGAN - Monet paintings. [link]
  • Improving CycleGAN - Monet paintings. [link]
  • Monet paintings (Berkeley) - TFRecords 256x256. [link]
  • Monet paintings - TFRecords 256x256. [link]
  • Keywords(Computer vision, Deep learning, GAN, CycleGAN, Art, Paintings)

SIIM-ISIC Melanoma Classification [Link for the complete project repository]

  • Melanoma Classification - SHAP model explained. [link]
  • Melanoma Classification - Augmentations EDA. [link]
  • Keywords(Computer vision, Deep learning, CNN, Transfer learning, Medical images, Melanoma detection, Object classification)


Tweet Sentiment Extraction [Link for the complete project repository]

  • Tweet Sentiment Extraction - EDA and baseline. [link]
  • Keywords(NLP, Deep learning, Transformer, Transfer learning, Question answering)


Jigsaw Multilingual Toxic Comment Classification [Link for the complete project repository]

  • Jigsaw - TPU optimized training loops. [link]
  • Jigsaw Classification - DistilBERT with TPU and TF. [link]
  • Keywords(NLP, Deep learning, Transformer, Transfer learning, TPU, Classification)


Flower Classification with TPUs

  • Flower Classification with TPUs - EDA and Baseline. [link]
  • Flower with TPUs - Advanced augmentation. [link]
  • Flower with TPUs K-Fold optimized training loops. [link]
  • Bug report unavailableerror socket closed. [link]
  • Keywords(Image, Deep learning, DenseNet, EfficientNet, Transfer Learning, Data augmentation, K-Fold, EDA, TPU, Tensorflow)


TensorFlow 2.0 Question Answering [Link for the complete project repository]

  • Google QUEST - EDA and USE Baseline. [link]
  • Keywords(NLP, Deep learning, Universal sentence encoder, transfer learning, EDA)


TensorFlow 2.0 Question Answering

  • Using TF 2.0 w/ Bert on NQ - (translated to TF2.0). [link]
  • TF2.0 Baseline w/ BERT - (translated to TF2.0). [link]
  • Keywords(NLP, Deep learning, BERT, transfer learning, Natural question, Open question)


Understanding Clouds from Satellite Images [Link for the complete project repository]

  • Understanding Clouds - EDA and Keras U-Net. [link]
  • Cloud Segmentation with utility scripts and Keras. [link]
  • cloud images segmentation utillity script. [link]
  • Keywords(Script, Keras, EDA, Deep Learning, Transfer learning, U-Net, Image segmentation)


APTOS 2019 Blindness Detection [Link for the complete project repository]

  • APTOS Blindness Detection - EDA and Keras ResNet50. [link]
  • Diabetic retinopathy - SHAP model explainability. [link]
  • (175th place) 5-Fold EfficientNetB5. [link]
  • Keywords(SHAP, Model explainability, EDA, Deep Learning, Image, Multi-class, Transfer learning, Health care, Classification)


Jigsaw Unintended Bias in Toxicity Classification [Link for the complete project repository]

  • Toxicity Bias - extensive EDA and Bi LSTM. [link]
  • Keywords(EDA, Deep Learning, Bidirectional GRU, Sequence, NLP, Classification)


iMet Collection 2019 - FGVC6 [Link for the complete project repository]

  • iMet Collection 2019 - EDA & Keras. [link]
  • iMet - Keras pretrained model as feature extractor. [link]
  • Keywords(Deep Learning, CNN, Transfer learning, Image, Multi-label classification)


Microsoft Malware Prediction [Link for the complete project repository]

  • EDA, data cleaning and feature visualization with LGB model classification part 1. [link]
  • EDA, data cleaning and feature visualization part 2. [link]
  • Keywords(LGBM, EDA, classification)


  • Instant Gratification - Deep Learning. [link]
  • Ensembling and evaluating Magic models. [link]
  • Keywords(EDA, Deep Learning, Emsembling, Classification)


  • Deep Learning Helping Navigate Robots. [link]
  • Keywords(EDA, Deep Learning, MLP, Time-series, Multi-class classification)


Google Analytics Customer Revenue Prediction [Link for the complete project repository]

  • EDA, feature engineering and selection, Keras deep learning revenue prediction.
  • EDA, feature engineering and selection, LGBM revenue prediction.
  • Keywords(Deep learning, MLP, LGBM, EDA, Regression)


New York City Taxi Fare Prediction [Link for the complete project repository]

  • Keras MLP Regression.
  • Tensorflow (estimator API) MLP Regression (end-to-end tensorflow project).
  • Keywords(Deep learning, MLP, Embedding, Regression)


Predict Future Sales

  • Model stacking (catboost, random forest and knn) time-series forecasting regression.
  • MLP using features encoded by a LSTM autoencoder and LSTM model time-series forecasting regression.
  • Keywords(Catboost, Random forest, KNN, Linear regression, Ensemble, Model stacking, MLP, LSTM, Autoencoder, Sklearn, Time-series, EDA, Regression)


Store Item Demand Forecasting Challenge [Link for the complete project repository]

  • Spark used to process the data and Keras deep learning recurrent neural networks for time series regression. [link]
  • MLP, CNN, LSTM, CNN-LSTM models and framing of time-series problem for forecasting regression. [link]
  • Keywords(Spark, Ml pipeline, Time-series preprocess, Deep learning, MLP, CNN, LSTM, CNN-LSTM, Time-series, Regression)


Titanic: Machine Learning from Disaster [Link for the complete project repository]

  • Naive Bayes classification [didactic code]. [link]
  • Tensorflow (core API) MLP classification. [link]
  • XGBoost classification and machine learning insights (Permutation Importance, Partial Dependence Plots, SHAP Values) [didactic code]. [link]
  • Keywords(Tensorflow, Sklearn, Deep learning, MLP, Naive Bayes, XGBoost , ML insights, Classification)


Quick, Draw! Doodle Recognition Challenge

  • Keras deep learning CNN classification.
  • Clean and pre process large csv files.
  • Competition EDA.
  • Converting drawings to numpy arrays.
  • Keywords(Deep learning, CNN, EDA, Classification)


Digit Recognizer

  • Deep learning CNN for image classification. [link]
  • Keywords(Deep learning, CNN, Image classification)


House Prices: Advanced Regression Techniques [Link for the complete project repository]

  • EDA, feature correlation, selection, engineering and visualization. [link]
  • Tensorflow (core API) MLP Regression. [link]
  • Keywords(Tensorflow, R, Deep learning, MLP, EDA, Regression)