** Welcome to my GitHub repo. **

I am a Data Scientist and I code in R, Python and Wolfram Mathematica. Here you will find some Machine Learning, Deep Learning, Natural Language Processing and Artificial Intelligence models I developed.

** Outputs of the models can be seen at my portfolio: ** https://drive.google.com/file/d/0B0RLknmL54khdjRQWVBKeTVxSHM/view?usp=sharing

** MNIST_HOT.5.FULL: ** is a solution for the MNIST dataset in Mathematica, with 96.51% accuracy, based on difference of pixels.

** Mathematica - Artificial Intelligence Simulating Interactions in Social Networks: ** is a model that simulates human interactions in a social network using cellular automata and agent-based modeling. Each agent has 3 possible choices for interation and a memory. The code has 14 pages with a big loop included in one line of code.

** Mathematica - Facial Recognition in Movement: ** This code operationalizes facial recognition in a downloaded YouTube video. The output is also a video with the result of face recognition (YouTube link of the output is included in code page)

** Mathematica - Monte Carlo Simulation: ** is an animated model of a Markov Chain Monte Carlo Simulation for autonomous driving. A video of the dynamic output was also generated and link for the YouTube video is included in code page.

** Mathematica - Social Network Surveillance: ** is a model that tracks individuals in a social network, tracks also his connections and future interactions.

Keras version used in models: keras==1.1.0 | LSTM 0.2

** Python - Autoencoder MNIST: ** is an autoencoder model for classification of images developed with Keras, for the MNIST dataset, with model Checkpoint as a callback to save weights.

** Python - Autoencoder for Text Classification: ** is an autoencoder model for classification of text made with Keras, also with model Checkpoint.

** Python - Deep Learning with Lasagne: ** is a deep neural network developed with Lasagne, where you can see values of weights in each layer, including bias.

** Python - Face Recognition: ** is a model using OpenCV to detect faces.

** Python - Image Extraction from Twitter: ** is a model that extracts pictures and their links from Twitter webpages, plotting with matplotlib.

** Python - Keras Convolutional Neural Network: ** is a CNN developed to classify the MNIST dataset with an accuracy greater than 99%.

** Python - Keras Deep Regressor: ** is a deep Neural Network for prediction of a continuous output made with Keras, learning rate scheduler according to derivative of error, random initial weights, with loss history.

** Python - Keras LSTM Network: ** is a Recurrent Neural Network (LSTM) to predict and generate text.

** Python - Keras Multi Layer Perceptron: ** is a MLP model, Neural Networks made with Keras with loss history, scheduled learning rate according to derivative of error for prediction and classification.

** Python - Machine Learning: ** is a Principal Components Analysis followed by a Linear Regression.

** Python - NLP Doc2Vec: ** is a Natural Language Processing model where I asked a Wikipedia webpage a question and 4 possible answers were semantically chosen from the tokenized and vectorized webpage, using KNN and cosine distance.

** Python - NLP Semantic Analysis: ** is a Natural Language Processing model that classifies a given sentence according to semantic similarity to other sentences, using cosine distance.

** Python - NLP Word2Vec: ** is a model developed from scratch to measure cosine similarity among words.

** Python - Reinforcement Learning: ** is a model based on simple rules and Game Theory where agents attitude change according to payoff achieved. Can be adapted for tit-for-tat strategy, always cooperate, always defeat and other strategies. Rewards were placed in the payoff matrix.

** Python - Social Networks: ** is a model that draws social networks configuration and connections.

** Python - Support Vector Machines: ** is a Machine Learning model that classifies the Iris dataset with SVM and plots it.

** Python - Theano Deep Learning: ** is a Neural Network with two hidden layers using Theano.

** R - Churn of Customers: ** is a model that uses a logistic regression associated with a threshold to predict which customers present the greater risk to be lost.

** R - Data Cleaning + Multinomial Regression: ** is a model that presents data cleaning and a multinomial regression using package nnet to classify customers according to their level of loyalty.

** R - Face Recognition: ** is a code to detect faces and objects in R.

** R - Geolocation Brazil: ** is a file for geo-spatial localization, brazilian map.

** R - Geolocation USA: ** is also a file for geo-spatial localization, USA map.

** R - Geolocation World: ** is a file for geo-spatial localization, world map, zoom available, customizable icons.

** R - Gradient Descent Logistic: ** is a model that performs a gradient descent to define a threshold for the sigmoid function in a Logistic Regression. Boosting was implemented and ROC curves compared.

** R - H2O Deep Learning: ** is a Neural Network model developed to predict recommendations and word-of-mouth advertising.

** R - Imbalanced classes ** is a model for employee churn, where features have no correlation with target variable and also there are imbalanced classes in the proportion 1/20. A logistic regression from scratch is applied, a hill climbing gradient is used to define the best threshold for the logistic function and after that, boosting was compared regarding AUC in a ROC plot.

** Logistic Regression + Gradient Descent + Boosting ** is a model where features have no correlation with target variable. Logistic Regression with Gradient Descent was applied, and then Boosting.

** R - MNIST: ** is a solution for the MNIST dataset, developed from scratch.

** R - Markov Chains: ** is a simple visualization of Markov Chains and probabilities associated.

** R - NeuralNet: ** is a Neural Network model developed to predict and classify word-of-mouth advertising.

** R - Ridge Regression: ** is a model with Ridge Regularization made from scratch to prevent overfitting.

** R - Deep Learning: ** is a Neural Network model with 2 hidden layers for prediction of a continuous variable.

Get A Weekly Email With Trending Projects For These Topics

No Spam. Unsubscribe easily at any time.

python (53,650)

deep-learning (3,922)

python3 (1,613)

nlp (1,078)

keras (762)

natural-language-processing (683)

rstats (308)

lstm (266)

python-3 (209)

face-recognition (172)

word2vec (114)

timeseries (103)

nlp-machine-learning (86)

mathematica (82)

autoencoder (79)

theano (75)

lstm-neural-networks (45)

time-series-analysis (42)

lasagne (19)