we would like to add the summary and notes on week 3 from course 2: Word embeddings with neural networks.

add summary and notes on course Autocorrect and Dynamic Programming

The gradient for biases defined does not work
grad_b1 = np.sum((1/batch_size)*np.dot(l1,x.T),axis=1,keepdims=True)
grad_b2 = np.sum((1/batch_size)*np.dot(yhat-y,h.T),axis=1,keepdims=True)

My suggestion to correct it:
grad_b1 = np.dot(l1,np.ones((batch_size,1)))/batch_size
grad_b2 = np.dot(yhat-y,np.ones((batch_size,1)))/batch_size

It is a similar thing to do but it gives a better understanding as well as, works for me!

In course 3 week 2 assignment,
data_generator = data_generator(batch_size, v_sentences, v_labels, vocab[''], True)
but in course 3 week 3 assignment,
data_generator = trax.supervised.inputs.add_loss_weights(
data_generator(batch_size, v_sentences, v_labels, vocab[''], True),
id_to_mask=vocab[''])
Why the data_generator didn't mask padding in the loss weights of data? @ijelliti

Can you add course 3 week2 ungraded labs？I didn't find the ungraded labs in course3 week2 directory. @ijelliti

In course 1, second week notes,
The concept of laplasian smoothing is discussed, but the result table is not there..
Please add following table to make it more intuitive.

Can you upload the lecture notebook?
NLP_C3_W1_lecture_notebook_introduction_to_trax.ipynb
NLP_C3_W2_lecture_notebook_Hidden_State_Activation.ipynb
NLP_C3_W2_lecture_notebook_Working_with_Jax_Numpy_and_Calculating_Perplexity.ipynb
NLP_C3_W2_lecture_notebook_Creating_a_GRU_model_Using_Trax.ipynb
NLP_C3_W3_lecture_notebook_Vanishing_Gradient.ipynb
NLP_C3_W4_lecture_notebook_Modified_Triplet_Loss.ipynb
NLP_C3_W4_lecture_notebook_Create_a_Siamese_Model_using_Trax.ipynb
NLP_C3_W4_lecture_notebook_Evaluate_a_Siamese_Model.ipynb

In ,the Accuracy of the Viterbi algorithm is 0.9528 while the expected is 0.9531
despite the difference is small 0.0003, the autograder consider it wrong accuracy.

output from the autograder:

Code Cell UNQ_C8: Function 'compute_accuracy' is incorrect. Check implementation.

Is it possible to use this code tot rain and chatbot in a different language?
Or are you using a pre-trained model that is pre-trained in English?

What changes are required to make it work in the German language?

Hi First, thank you very much for sharing your solutions.

I noticed in some notebooks we need to call functions defined in the module/script called utils. I am wondering if we can find that script anywhere in the repo? It seems it can not be found if we run the notebook on a local machine.

Thanks again!

we would like to add the summary and notes on week 3 from course 2: Autocomplete and Language Models

at exercise 10 perplexity: the function five the following output

Perplexity for first train sample: 2.8944
Perplexity for test sample: 3.4760

while the Expected Output is:

Perplexity for first train sample: 2.8040
Perplexity for test sample: 3.9654

We would like to share the summary and notes on week 4 of Course 1 : Word Translation

Just like previous weeks of the first course. we would like to share the summary and notes on week 3: Word Embeddings

following week 1 course notes, we continue to add the same to the next weeks of courses.
this issue is to continue share notes on naive Bayes from Week 2 Course1.

In 1 - Natural Language Processing with Classification and Vector Spaces - Labs - Week1 - NLP_C1_W1_lecture_nb_03.ipynb.

There is no 'logistic_features.csv'.

Course 3 started yesterday, can you upload the init code? Thank you!

In C1_W2_Assignment , the utlis.py doesn't have the function lookup.

On the 2nd course of the specialization on the assignment of the week 4 there is a mistake on the back-propagation calculation on the section 'UNQ_C4'

You calculate the gradient of biases wrong. You should change it as follows:

grad_b1 = (1/batch_size)*np.dot(l1,ones_array.T).reshape(-1,1)

grad_b2 = (1/batch_size)*np.dot(yhat-y,ones_array.T).reshape(-1,1)

where

ones_array = np.ones(batch_size)

With that way it is the same as the slides in the course. You can check your own notes too on this github repository of how gradient of bias should be calculated.

Course 3 started last week, can you upload the lecture notebook ? Thank you

We would like to share the summary and notes on week 3: Part of Speech Tagging and Hidden Markov Models

Describe the bug
In NLP_C1_W4_lecture_nb_01 there is no utils_nb.py

Expected behavior
add the utils_nb.py

can you upload the ungraded notebook in course4 week4？the Ungraded Lab: Reformer LSH notebook and the Ungraded Lab: Revnet notebook @ijelliti

Shouldn't be there a function confidence_ellipse in the utils.py in the lab for Course1, week2?

The line "from utils import confidence_ellipse" would return an error otherwise

Sorry, I can't find this csv.
'bayes_features.csv'.