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
Poetry is just the evidence of life. If your life is burning well, poetry is just the ashLeonard Cohenburn somethingmore quotes

art: fun



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


art + science

Bloomberg Businessweek Design Conference — San Francisco, 2013

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Design loves science and science loves design, but doesn't always know it. (Bloomberg Businessweek)

science design

Together with Alberto Cairo, I presented at the Bloomberg Businessweek Design Conference (highlights) on the topic of design and communication in the sciences.

Alberto, as the journalist, motivated why communication should include access to detail through an engaging narrative. He made the distinction between the specialist (heavy on detail) and the communicator (focus on narrative) and emphasized that the distinction is artificial, though often played out (watch video).

I, as the scientist, underscored the importance of clear communication between scientists. As the specialists, they are often very poor communicators. Pick up any science journal and you'll quickly discover that scientists either aren't good at telling stories or are discouraged to do so by the medium. The consequence is the same: papers read like a wall of text. TL;DR anyone? The quality of visual communication in general ranges from muddled to abysmal (watch video).

We need more leaders in the field, such as Nature Publishing Group, to reward and emphasize good visual communication (e.g. Nature Cancer Review 2013 Figure Calendar).

Our presentations concluded with a 15 minute moderated discussion with Sam Grobart, senior Businesssweek writer. Everyone got a little cheeky. Good fun.

presentation video

Watch: my presentation, conversation with Alberto Cairo, moderated by Sam Grobart. (Bloomberg TV), Albert Cairo's presentation.

presentation slides

This was a lightning 7 minute talk. I did more planning about what to say than I usually do, given that there was virtually no opportunity for any kind of backtracking, and include a running narrative below each slide.

Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
1/32

download presentation

My slides are available as PDF, keynote (zipped) or Quicktime. The format is 16:9 HD.

Bloomberg Businessweek Design Issue

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
The reality of redesign is disruptive. How can we pursue new ideas and opportunities without leaving consumers confused or angry? Businessweek puts that question to some of the world's most accomplished designers. (Bloomberg Businessweek Design Issue)

On 28 Jan 2013, Bloomberg Businessweek Design Issue will capture the ideas from the conference and the personalities that generated them.

During the conference, each talk was captured in a series of sketches by Tom Wujec: my talk sketch and moderated discussion sketch.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Date completed: ongoing — an accurate assessment of the state of the visual communication field in science. (read article)
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

Happy 2016 `\pi` Approximation, roughly speaking

Sun 24-07-2016

Today is the day and it's hardly an approximation. In fact, `22/7` is 20% more accurate of a representation of `\pi` than `3.14`!

Time to celebrate, graphically. This year I do so with perfect packing of circles that embody the approximation.

By warping the circle by 8% along one axis, we can create a shape whose ratio of circumference to diameter, taken as twice the average radius, is 22/7.

If you prefer something more accurate, check out art from previous `\pi` days: 2013 `\pi` Day and 2014 `\pi` Day, 2015 `\pi` Day, and 2016 `\pi` Day.