Lesson 2 - Official Topic

Thanks for the information and yes I will take a look at that in the docs.

For those of you that (like me) had image failures on .webp images, you can install the webp software into python by doing:

pip install webp

Then the image failures go away

2 Likes

Jeremy’s public Covid19 video

Just a small correction. Zuzana Čaputová is a president, not prime minister.

BTW she was an inspiration for Czech masks for all movement (she wore a mask and encouraging it in a moment when Czech government ignored it and there were no masks for sale in Czechia). You can find many pictures on social media contrasting her, wearing a mask, to Czech top officials. (To give you some analogy, Zuzana Čaputová is for Czech people what Justin Trudeau is for Americans.)

Hi,

The free article referenced below is written by people who more or less wrote ‘the Bible of epidemiology’ (Modern Epidemiology by Rothmann, Greenland and Lash):

A quote:
" A significant test result ( P 0.05) means that the test hypothesis is false or should be rejected. No! A small P value simply flags the data as being unusual if all the assumptions used to compute it (including the test hypothesis) were correct; it may be small because there was a large random error or because some assumption other than the test hypothesis was violated (for example, the assumption that this P value was not selected for presentation because it was below 0.05). P ≤ 0.05 only means that a discrepancy from the hypothesis prediction (e.g., no difference between treatment groups) would be as large or larger than that observed no more than 5 % of the time if only chance were creating the discrepancy (as opposed to a violation of the test hypothesis or a mistaken assumption)."

And yes, that’s probably the book. It demonstrate many examples with R (the programming language).

3 Likes

We conducted lesson 2’s re-stream on Wednesday 7pm IST from our Bengaluru, India virtual study group , happy to inform that 55 people attended the session. We got a couple of questions from the participants:

  1. In the Covid-19 paper, in Fig 4, R is plotted between 1 and 2.5. I am slightly confused about the definition of R. Can R be less than 1? Isn’t it desirable to bring R to <1?
  2. How do we use new pre-trained model from model zoo is there article on this?

Thanks.

1 Like

i.e. Can a model be reversed engineered? It would be easier to reverse if the model’s architecture is known, if the training dataset is available, if the dataset is known to be derived from a known dataset. For example, you create a stock trading model and license it without distributing the dataset. It might be possible to create a new dataset from the model. The fact that stock prices are publicly known might make it easier to reconstruct the original dataset forming the basis of an adversarial attack.

2 Likes

Hi Kofi,

I ran into the same situation. I found a link to the keys in the email sent after making an Azure account.

1 Like

great thanks @DanielLam

Please I want to know if voila works in paperspace. I’m trying to use it but replacing notebook with voila/render takes me to the homepage. Thanks
@dkobran @muellerzr @init_27

1 Like

I am trying to use binder to serve the bear classifier app; but i am getting this error. Any suggestions would be helpful

2 Likes

Hi, I got the webp image failures also, but I thought they wanted to show us how to unlink the failed images, so I didn’t install but web software. Thanks for the tip…I didn’t know you could do that to get more images in the model.

I wrote a Jupyter notebook How can we determine a p-value for an experimental result? in which I carry out the experiment Jeremy suggested in Lesson 2 in relation to the discussion of Figure 1(a) of the paper "High Temperature and High Humidity Reduce the Transmission of COVID-19"

The result may surprise you!

4 Likes

Regarding COVID-19, this just came out:

(NOTE: very strong content)

I think it is worth posting, because today I went to retrieve my groceries and there were people going around like nothing is happening… wake up, this is real!

I have many friends in Bergamo…

Hi, i have a dataset which is organised like in the pic attached. How to create ImageDataBunch from this. The tree for ‘a’ is repeated for the rest
Screenshot 2020-03-28 at 7.35.49 PM

I have been reading the api docs for quite some time but was unable to get it

Does anybody know some bibliography that expands on the topics covered in the article “Designing great data products”?

Are a,b,c,d the classes you want to predict? What’s x,y?

x,y are the ones i need to predict. a,b,c,d are diff forms of x,y. Like if x & y are some image then a,b,c,d are like sketched, painted… of the image. does that make sense

In that case it should be enough to do

dblock = DataBlock(blocks=(ImageBlock, CategoryBlock()),
                           get_items=get_image_files,
                           splitter= GrandparentSplitter(train_name='train', valid_name='validation'),
                           get_y=parent_label,
                           )
dls =ImageDataLoaders.from_dblock(dblock, path, path=path)

This will merge all training and validation samples across a,b,c,d.

1 Like

Are we supposed to deploy our classifier notebook that we created? For the step that says to add the notebook to your github repository, how would you do that if using paperspace?