cs224d Problem Set 1 job __cs224

Welcome reprint, reprint indicate the origin:

http://blog.csdn.net/neighborhoodguo/article/details/47071597
The detailed derivation process of computational problems in 1.assignment1

The negative sampling in the programming question is similar to the derivation process of the third question (a), so there is no deduction

4. (a) because over-fitting'll make model have a poor generalized error and overfit the training set. In order to improve our model's accuracy, we should introduce regularization to avoid over-fitting.

(b) This is a "V" curve shape, since less regularization'll leads to under-fitting, i.e. the model doesn ' t catch enough featu Res, more regularization'll leads to over-fitting, i.e. the model catch noise in the training set this is also bad for OU R model. The middle the complexity of our model is good, this is a trade-off between catching more features and less noise.

2. Procedure GitHub

The following is my cs224d GitHub address, have to refer to friends can refer to. But it's not very good, don't spray me.

Https://github.com/NeighborhoodWang/CS224D-problem-set/tree/master/Psets/Pset1/assignment1

Write a program when you encounter several problems, recorded:

1. In Python, I like to represent vectors in one line, which makes it a bit inappropriate for me to use MATLAB.

2. The way to assign a matrix or vector to other variables is to use the x.copy (), otherwise it just assigns the pointer to the past.

3.gradient Check to use the high precision that is, both sides of the Eplison minus the one (this is killing me)