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.

My Kaggle profile

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 [Link for the complete project repository]

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

---

KaggleCareerCon2019 [Link for the complete project repository]

• 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)