PlayGround

Basically my own Area51. Here I post my first attempts with new methods and datasets. Some of these experiments are embarrassing incomplete, baby steps, but an overall valuable learning experience.

On Progress

• 20170412_LSTM-pubmed-titles. Generating Titles for scientific Papers. Playing with LSTM, try to learn how to generate titles for pubmed papers. (playing with Keras) ToDo write a sampler, train the model with more data. Add embeddings to code the words.

• 20171213_Interview_prediction. Predict candidates attendance to interviews, Tried LogisticRegression, RandomForest, and Bayesian models, I did not get a mindblowing accuracy so far. ToDo Work a bit more in the feature engineering, try some other approach to optimize model.

• 20180213_Items_Prices_predict. Prices Predictions. An example of a regression model, trying to predict prices using Keras. ToDo Work a bit more on features, try other models.

• 20180308_Monkeys_class. Identifying Monkeys. Learning pythorch. Build a classification model for monkey images. (Playing with pyTorch) ToDo try transferring learning, resnet50?

• 20180120_VAE_cancer_clusters. Better clustering of cancer patients using Variational Autoencoders. (Pytorch) ToDo Improve validation.

• 20180213_Items_Precies_predict. Prices Predictions. a example of a regression model, trying to predict prices using Keras. ToDo Work a bit more on features, try other models.

Almost Done

• 20180520_TPOT_with_Titanic. Testing TPOT, an automated Machine learning tool that optimizes machine learning pipelines using genetic programming. Using Kaggle competition on prediction of survival for passengers of the Titanic as a dummy example. ToDo I could work more on features, however, this is pretty much done, I like TPOT!

• 20180725_Smartlibrary_generator. Generator of peptides enforcing a concreate secondary structre using RNN (pytorch)

The Data Explorer Code

Here you will find all the code and Jupyter notebooks I used for the entries of The Data Explorer.

• Face Generator (2017/09/10). Using GANs to create a photorealistinc Face generator. Training Video (Playing with Keras)

• Cooking Word2Vec (2017/03/19). Playing eith distributed word vectors (word2vec) for clustering cooking reciptes. next, build a predictor using the vectors as features. (playing with Gensim)