Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca - contact me Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca on Twitter Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca - Lumondo Photography Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca - Pi Art Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca - Hilbertonians - Creatures on the Hilbert Curve
Trance opera—Spente le Stellebe dramaticmore quotes

satire: beautiful


UCD Computational and Molecular Biology Symposium, Dublin, Ireland. 1-2 Dec 2016.


fun + amusement

Dummer — Like Nothing Else


The Hummer font is a slightly modified Antique Olive Nord. The Like Nothing Else tag line is Trade Gothic. Both have character widths increased to 110-120% and individually adjusted kerning. Get the Illustrator CS5 file for both logos.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Hummer logo. (EPS, PNG)
Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer logo. (EPS, PNG)
Download high-resolution images.

This project might give you the impression that I don't like Hummers. You'd be right.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
It could be worse. But not by much. (zoom)
Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
It could be worse. But not by much. (zoom)
Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
It could be worse. But not by much. (zoom)

update

The Maurauder. Over 25,000 lb — five times what an H3 weighs. Enough said.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
There is always someone with a bigger one. (Manufacturer's page.)

Dummer - Like Nothing Else

Hummers are a cultural equivalent of a toxic warning label and have the same effect on me as bug spray on mosquitoes.

I am not the first one to satirize this automotive aberration, so there's some hope.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like Nothing Else. (New York Times — Laugh Lines)

GM's advertisement images require no modification for the satire, which makes it all that much better.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dumb and Dumber. (New York Times — Laugh Lines)

I could have just as well used the Lincoln Navigator or Cadillac Escalade, but they don't embody the superlative like the Hummer.

The Hummer brand proved itself to be aesthetically, rationally and economically unsustainable and collapsed after a failed attempt to sell it to China. There continues to be a robust market for used Hummers. Let the farce continue.

I'm hated

It delights me that this project produced my first hate mail.

I want to meet Doug and give him a hug for adding another dimension to this project.

I'm loved

The images got picked up by the New York Times laughlines blog, which drew a couple of fan mails.

But neither made me feel as good as Doug's email.

Dummer Images

Dummer. Like nothing else. A pretty good Hummer satire. / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like nothing else. (zoom)
Dummer. Like nothing else. A pretty good Hummer satire. / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like nothing else. (zoom)
Dummer. Like nothing else. A pretty good Hummer satire. / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like nothing else. (zoom)
Dummer. Like nothing else. A pretty good Hummer satire. / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like nothing else. (zoom)
Dummer. Like nothing else. A pretty good Hummer satire. / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like nothing else. (zoom)
Dummer. Like nothing else. A pretty good Hummer satire. / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like nothing else. (zoom)
Dummer. Like nothing else. A pretty good Hummer satire. / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like nothing else. (zoom)
Dummer. Like nothing else. A pretty good Hummer satire. / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like nothing else. (zoom)
Dummer. Like nothing else. A pretty good Hummer satire. / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Dummer. Like nothing else. (zoom)
VIEW ALL

news + thoughts

Intuitive Design

Thu 03-11-2016

Appeal to intuition when designing with value judgments in mind.

Figure clarity and concision are improved when the selection of shapes and colors is grounded in the Gestalt principles, which describe how we visually perceive and organize information.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
One of the Gestalt principles tells us that the magenta and green shapes will be perceived as as two groups, overriding the fact that the shapes within the group might be different. What the principle does not tell us is how the reader is likely to value each group. (read)

The Gestalt principles are value free. For example, they tell us how we group objects but do not speak to any meaning that we might intuitively infer from visual characteristics.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Nature Methods Points of View column: Intuitive Design. (read)

This month, we discuss how appealing to such intuitions—related to shapes, colors and spatial orientation— can help us add information to a figure as well as anticipate and encourage useful interpretations.

Krzywinski, M. (2016) Points of View: Intuitive Design. Nature Methods 13:895.

...more about the Points of View column

Regularization

Fri 04-11-2016

Constraining the magnitude of parameters of a model can control its complexity.

This month we continue our discussion about model selection and evaluation and address how to choose a model that avoids both overfitting and underfitting.

Ideally, we want to avoid having either an underfitted model, which is usually a poor fit to the training data, or an overfitted model, which is a good fit to the training data but not to new data.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Nature Methods Points of Significance column: Regularization (read)

Regularization is a process that penalizes the magnitude of model parameters. This is done by not only minimizing the SSE, `\mathrm{SSE} = \sum_i (y_i - \hat{y}_i)^2 `, as is done normally in a fit, but adding to this minimized quantity the sum of the mode's squared parameters, `\mathrm{SSE} + \lambda \sum_i \hat{\beta}^2_i`.

Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Regularization. Nature Methods 13:803-804.

Background reading

Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Model Selection and Overfitting. Nature Methods 13:703-704.

Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Classifier evaluation. Nature Methods 13:603-604.

Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Logistic regression. Nature Methods 13:541-542.

...more about the Points of Significance column

Model Selection and Overfitting

Fri 04-11-2016

With four parameters I can fit an elephant and with five I can make him wiggle his trunk. —John von Neumann.

By increasing the complexity of a model, it is easy to make it fit to data perfectly. Does this mean that the model is perfectly suitable? No.

When a model has a relatively large number of parameters, it is likely to be influenced by the noise in the data, which varies across observations, as much as any underlying trend, which remains the same. Such a model is overfitted—it matches training data well but does not generalize to new observations.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Nature Methods Points of Significance column: Model Selection and Overfitting (read)

We discuss the use of training, validation and testing data sets and how they can be used, with methods such as cross-validation, to avoid overfitting.

Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Model Selection and Overfitting. Nature Methods 13:703-704.

Background reading

Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Classifier evaluation. Nature Methods 13:603-604.

Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Logistic regression. Nature Methods 13:541-542.

...more about the Points of Significance column

Classifier Evaluation

Tue 13-09-2016

It is important to understand both what a classification metric expresses and what it hides.

We examine various metrics use to assess the performance of a classifier. We show that a single metric is insufficient to capture performance—for any metric, a variety of scenarios yield the same value.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Nature Methods Points of Significance column: Classifier Evaluation (read)

We also discuss ROC and AUC curves and how their interpretation changes based on class balance.

Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Classifier evaluation. Nature Methods 13:603-604.

Background reading

Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Logistic regression. Nature Methods 13:541-542.

...more about the Points of Significance column