udacity-deep_learning

Jupyter notebooks (with answers) used during the Google Deep Learning MOOC

The goal of this MOOC is to complete all the assignments in the "tensorflow/examples/udacity" directory of tensorflow

MOOC url: https://www.udacity.com/course/deep-learning--ud730

Status: DONE!

Assigment 1: logistic regression

nearly DONE

• I have not created the sanitized dataset. (Anyway the results seem good with messy data)

BONUS:

• Comparison of logistic regression results with 2 other machine learning algorithms from scikit learn: gaussian Naive Bayes and random forest regressor

Assigment 2: basic neural network

DONE

Assigment 3: regularization

DONE

Assigment 4: convolutions

DONE (best performance with the NotMNIST dataset: 96.1%)

Assigment 5: word2vec

DONE (CBOW model in word2vec)

Assigment 6: recurrent neural networks with LSTM (and GRU) cells

• Problem 1: DONE
• Problem 2: DONE
• Problem 3: DONE (sequence-to-sequence models)

BONUS for problem 2:

• implement a GRU cell instead of an LSTM cell
• mix LSTM and GRU cells at the same time
• the GRU equations are taken from this great blog post on RNNs: http://colah.github.io/posts/2015-08-Understanding-LSTMs/