I started Standford's machine learning course but after getting stuck in one assignment (ex4, Week5), it fell of the table due to lack of time and focus.

I will use this Hack Week to make some progress on it.

Day 1

• Studied back-propagation algorithm again, starting Week 5 from scratch

Day 2

• Focused in the homework itself, understanding what I did last time.
• Cross checked vector dimensions, to understand better
• Simplified the code by vectorizing the second part the whole thing
• Solved a mystery with dimensions realizing that in order to calculate δ2, I need to ignore the first column of Θ2 (bias)

• Some progress

  Part Name | Score | Feedback --------- | ----- | -------- Feedforward and Cost Function | 30 / 30 | Nice work! Regularized Cost Function | 15 / 15 | Nice work! Sigmoid Gradient | 5 / 5 | Nice work! Neural Network Gradient (Backpropagation) | 40 / 40 | Nice work! Regularized Gradient | 0 / 10 | -------------------------------- | 90 / 100 |

• Working now on the regularized gradient. Homework additional material was helpful to understand the restriction of j=0

  Part Name | Score | Feedback --------- | ----- | -------- Feedforward and Cost Function | 30 / 30 | Nice work! Regularized Cost Function | 15 / 15 | Nice work! Sigmoid Gradient | 5 / 5 | Nice work! Neural Network Gradient (Backpropagation) | 40 / 40 | Nice work! Regularized Gradient | 10 / 10 | Nice work! -------------------------------- | 100 / 100 |

• Continued with Week 6

  • Model Selection
  • Ways to diagnose bias vs variance by separating data into cross-validation and test sets

Day 3

• Used on another Hackweek project

Day 4

• Continued with model selection and bias/variance vs other parameters

Day 5

• Finished Week 6 material. Debugging models and strategies to attack high bias or high variance
• Quiz for Week 6. 100%
• Started homework (ex5)
• Implemented Regularized Linear Regression Cost Function and Regularized Linear Regression Gradient