Technique of the day

How do you generate images from low dimensional vectors?

How do you go from low dimensional data (like a vector of size 100) to higher dimensional data (like an image of size 200x200)?

In many Generative AI applications, you’d need to do this. So how can you do it?

In 2D, you have 2 options:

• Conv2DTranspose.
• UpSampling2D + Conv2D.

What’s the difference between the two options?

When you use Conv2DTranspose, you’re doing upsampling by adding zeros vertically and horizontally.

When you use UpSampling2D + Conv2D, you’re doing upsampling by duplicating rows and columns.

When to use which?

There is no one answer. It really depends on your experimentation.

One drawback though for Conv2DTranspose, is that sometimes the output generated image can have some checkerboard artifacts like the image below.

Tip of the day

A recurrent problem with conda and how to solve it!

Most of us in the field of machine learning are using conda. But most of us face this problem at some point.

How do you create virtual environments on your local disk in a directory other than the default "Anaconda3/envs"?

Since we usually end up installing Anaconda in the default path during the installation process, it ends up being installed in the C drive.

This can quickly fill up the C disk.

So if you decide to create virtual environments and save them in another drive, how can you do that?

Well, you can achieve this by using the argument --𝐩𝐫𝐞𝐟𝐢𝐱.

To install a new virtual environment in a directory inside D drive for example, you can do this:

𝐜𝐨𝐧𝐝𝐚 𝐜𝐫𝐞𝐚𝐭𝐞 --𝐩𝐫𝐞𝐟𝐢𝐱=𝐃:\𝐩𝐚𝐭𝐡\...\𝐲𝐨𝐮𝐫𝐄𝐧𝐯𝐍𝐚𝐦𝐞 𝐩𝐲𝐭𝐡𝐨𝐧=𝐱.𝐱

One thing you should know is, conda only keeps track of the environments included in the folder envs inside the anaconda folder.

So, the next time you need to activate your new virtual environment, move to the folder where you created it and activate it with:

𝐜𝐨𝐧𝐝𝐚 𝐚𝐜𝐭𝐢𝐯𝐚𝐭𝐞 𝐲𝐨𝐮𝐫𝐄𝐧𝐯𝐍𝐚𝐦𝐞

Tweet of the day

ChatGPT is now available on Azure!

Microsoft Azure @Azure Words you’ve been waiting for—ChatGPT is now available on Azure OpenAI. Learn more → msft.it/60145s7by #AzureOpenAI #ChatGPT pic.twitter.com/dXgk840cL6 March 9th 2023 76 Retweets 218 Likes

​

ChatGPT understanding what full captioning means!

HIT THE INBOX OF MORE THAN 1500 MACHINE LEARNING PRACTITIONERS

Advertise with AIFEE to get your brand in front of hundreds of readers who are eager to start or accelerate their machine learning careers by learning about the latest and greatest in machine learning from: books, courses or online communities.

Many of our readers are also looking for machine learning related jobs.

​Get in touch today.

What'd you think of today's edition?

• ​🤖🤖🤖🤖🤖 Absolutely Great!​
• ​🤖🤖🤖 Meh, do better​
• ​🤖 You didn't bring the heat​