Tag: paper

AutoML-Zero: Evolving Machine Learning Algorithms From Scratch

AutoML-Zero: Evolving Machine Learning Algorithms From Scratch

Google Brain researchers published this amazing paper, with accompanying GIF where they show the true power of AutoML.

AutoML stands for automated machine learning, and basically refers to an algorithm autonomously building the best machine learning model for a given problem.

This task of selecting the best ML model is difficult as it is. There are many different ML algorithms to choose from, and each of these has many different settings ([hyper]parameters) you can change to optimalize the model’s predictions.

For instance, let’s look at one specific ML algorithm: the neural network. Not only can we try out millions of different neural network architectures (ways in which the nodes and lyers of a network are connected), but each of these we can test with different loss functions, learning rates, dropout rates, et cetera. And this is only one algorithm!

In their new paper, the Google Brain scholars display how they managed to automatically discover complete machine learning algorithms just using basic mathematical operations as building blocks. Using evolutionary principles, they have developed an AutoML framework that tailors its own algorithms and architectures to best fit the data and problem at hand.

This is AI research at its finest, and the results are truly remarkable!

GIF for the interpretation of the best evolved algorithm

You can read the full paper open access here: https://arxiv.org/abs/2003.03384 (quick download link)

The original code is posted here on github: github.com/google-research/google-research/tree/master/automl_zero#automl-zero

GIF for the experiment progress
Tidy Missing Data Handling

Tidy Missing Data Handling

A recent open access paper by Nicholas Tierney and Dianne Cook — professors at Monash University — deals with simpler handling, exploring, and imputation of missing values in data.They present new methodology building upon tidy data principles, with a goal to integrating missing value handling as an integral part of data analysis workflows. New data structures are defined (like the nabular) along with new functions to perform common operations (like gg_miss_case).

These new methods have bundled among others in the R packages naniar and visdat, which I highly recommend you check out. To put in the author’s own words:

The naniar and visdat packages build on existing tidy tools and strike a compromise between automation and control that makes analysis efficient, readable, but not overly complex. Each tool has clear intent and effects – plotting or generating data or augmenting data in some way. This reduces repetition and typing for the user, making exploration of missing values easier as they follow consistent rules with a declarative interface.

The below showcases some of the highly informational visuals you can easily generate with naniar‘s nabulars and the associated functionalities.

For instance, these heatmap visualizations of missing data for the airquality dataset. (A) represents the default output and (B) is ordered by clustering on rows and columns. You can see there are only missings in ozone and solar radiation, and there appears to be some structure to their missingness.


Another example is this upset plot of the patterns of missingness in the airquality dataset. Only Ozone and Solar.R have missing values, and Ozone has the most missing values. There are 2 cases where both Solar.R and Ozone have missing values.b.JPG

You can also generate a histogram using nabular data in order to show the values and missings in Ozone. Values are imputed below the range to show the number of missings in Ozone and colored according to missingness of ozone (‘Ozone_NA‘). This displays directly that there are approximately 35-40 missings in Ozone.

c.JPGAlternatively, scatterplots can be easily generated. Displaying missings at 10 percent below the minimum of the airquality dataset. Scatterplots of ozone and solar radiation (A), and ozone and temperature (B). These plots demonstrate that there are missings in ozone and solar radiation, but not in temperature.d.JPG

Finally, this parallel coordinate plot displays the missing values imputed 10% below range for the oceanbuoys dataset. Values are colored by missingness of humidity. Humidity is missing for low air and sea temperatures, and is missing for one year and one location.


Please do check out the original open access paper and the CRAN vignettes associated with the packages!


Generating images from scratch: Parallel Multiscale Autoregressive Density Estimation

Generating images from scratch: Parallel Multiscale Autoregressive Density Estimation

A while ago, I blogged about this new algorithm, pix2code, which takes in pictures of graphical user interfaces and outputs the underlying code. Today, I discovered another fantastic algorithm, by Scott Reed and his colleagues at Google Deepmind. txt2pix would be a catchy name for this algorithm, as it can take in a fairly complex sentence (e.g., “a grey bird with a black head, orange eyes, and a yellow beak“) and generate a completely new and unique image based on its content. In their recently published paper, they elaborate on the algorithms inner workings.

An example of the training and generation process reported in the paper

Scott and his team have been working on this project for quite some time. The early version of the algorithm generated an image one pixel at a time, but it had difficulties generating large or high-quality images. After picking a starting pixel to generate, any consecutively generated pixel the algorithm generates needs to align with its neighbours. For example, if pixel A is the first pixel in the generation of the yellow beak of a bird, any pixels that are created in the neighbourhood of that pixel should take into account that pixel A is trying to visualize a yellow beak, and behave accordingly: either continuing the beak, or ending the beak and starting on another element of the image.

The problem with such an iterative approach (i.e., pixel by pixel) is that it can take a very long time for a computer to generate an image. Considering that a fairly small image, say 256 by 256 pixels, already contains 65.536 pixels, each of which needs to be generated while considering all its neighbours and keeping in mind the bigger picture. In the most recent, updated version of the algorithm, Scott and his team have allowed the generation of multiple unrelated pixels simultaneously at different ‘zones’ of the image. Hence the Parallel in Parallel Multiscale Autoregressive Density Estimation. With this parallel approach, the algorithm can now generate the pixels representing the yellow beak in one area of the image, while simultaneously generating pixels for the bird’s wings and the branch it’s sitting on at different sections of the image. This speeds up the process quite extensively, demanding less computation time, thus allowing for quicker image generation.

I can definitely recommend that you check out Scott Reeds’ twitter feed for some amazing animated GIFs of the generation process:

If you want to know more details behind the algorithm but do not fancy reading the entire paper, I recommend this short explanation video by Károly Zsolnai-Fehér (what a name!) of Two Minute papers: