This week in: Today I learned (TIL)
I’ve been keeping a running list of some of the interesting things that I learned these last couple of weeks. Hopefully you steal a trick or two from me here!
2021-05-04
Today I learned that this… well… works. I learned how to create a value based on another value from my redux store. It’s ugly, but it works. I’m not proud of this but it gets the job done.
const messages = useSelector((state) => state.messages);
const found = messages.find((message) => message.date == todayString);
const todaysMessage = found ? found : undefined;
const allThreeDone = found ? todaysMessage.answer1 != '' && todaysMessage.answer2 != '' && todaysMessage.answer3 != '' : undefined;
Today I learned how to count the amount of trues in an array.
[false,false,true,false,true].filter(x => x).length // returns 2
Today I learned how to add opacity to an image in react native.
<Image opacity={0.2} />
Today I learned how to copy your public key to your clipboard.
pbcopy < ~/.ssh/id_rsa.pub
Today I learned how to check what shell you are using.
echo $SHELL
Today I learned how to download node with wget if you are ssh’d into a server.
wget https://nodejs.org/dist/v10.5.0/node-v10.5.0-linux-x64.tar.xz
2021-05-05
Today I learned how to plot a confusion matrix using ConfusionMatrixDisplay in matplotlib.
mat = confusion_matrix(y_true=eval_labels, y_pred=predicted_eval_labels)
ConfusionMatrixDisplay(mat, display_labels=class_names).plot()
Today I learned about this neat mapping trick. If you have a list and you want to turn it into a dictionary mapping, simply turn it into a numpy array.
class_names = np.array(["apple", "pear", "banana"])
class_names[[1, 0]] # ["pear", "apple"]
–
Today I learned that if you want to save whatever figure you just make a new figure using plt.subplots() and pass through the ax object.
fig, ax = plt.subplots()
confusion_mat = confusion_matrix(y_true=labels, y_pred=predicted_labels)
ConfusionMatrixDisplay(confusion_mat, display_labels=class_names).plot(xticks_rotation='vertical', ax=ax)
fig.savefig("tmp.png")
Today I learned that, if your plots are showing correctly but savefig is clipping your axes, you should use the bbox_inches=tight setting.
Today I learned how to remove unused axes from images in a subplot.
def show_img_prediction_grid(imgs, labels, predicted_labels, class_names):
"""Shows a grid of images with predictions as coloured borders."""
n = int(np.ceil(len(imgs)**.5))
fig , axs = plt.subplots(n, n, figsize=(3 * n, 3 * n))
for i, (img, label, predicted_label) in enumerate(zip(imgs, labels, predicted_labels)):
title = f"Predicted: {class_names[predicted_label]}"
show_img_prediction(img, label, predicted_label, class_names, ax=axs[i // n][i % n], title=title)
# Delete empty axes
# This works as follows: square n^2 which gives us the grid, subtract
# the total amount of images we have, this gives us the number of last images that
# are empty and we need to delete from the axes
# only works if we don't have a square
grid_size = n**2
if grid_size % len(imgs) != 0:
for i in np.arange(grid_size)[-(grid_size - len(imgs)):]:
fig.delaxes(axs.flatten()[i])
Today I learned that order matters when doing logical and (&) in Python. a & b == c is not the same as a & (b == c). Duh. This is very basic but I forgot some parentheses and it stumped me for a while.
2021-05-06
Today I learned that there is an ideal time to review your notes.
SuperMemo is based on the insight that there is an ideal moment to practice what you’ve learned. Practice too soon and you waste your time. Practice too late and you’ve forgotten the material and have to relearn it. The right time to practice is just at the moment you’re about to forget.
Today I learned that the first question you should always ask yourself during debugging is: “Can I reproduce this bug?”
2021-05-09
Today I learned that the Rectified Linear Unit (ReLU) activation function solves the vanishing gradient problem. I should know this, but I forgot.
Today I learned that Tensorflow.js stores the model as a graph, but loses some precision due to different data structures. Or so I’ve been told.
Today I learned you can use $$ ... $$ to add LaTeX on this blog. I completely forgot about this hah! I actually wrote a blog post about this a while back.
2021-05-17
Today I learned that working in 3 sprints of 25 minutes works really well for me. What works even better is if you set clear goals for those pomodoros. Usually you finish your goal with some time left. You can do a lot in 25 minutes!
Today I learned that you can check whether an array is empty by checking the length: array.length == 0 .
Today I learned how to add a line break in a <Text> component.
<Text
style={{
fontStyle: 'italic',
color: 'grey',
}}
>{`
It's quite empty here...
try adding something in the journal screen
`}
</Text>
Today I learned that you can check if a string is empty using triple quotes: str === "".
Today I learned how to map an array to an object.
const data = [
{id: 1, country: 'Germany'},
{id: 2, country: 'Austria'},
{id: 3, country: 'Switzerland'}
];
const dictionary = Object.assign({}, ...data.map(x => ({[x.id]: x.country})));
// {1: "Germany", 2: "Austria", 3: "Switzerland"}
Conclusion
I hope you scanned over these TILs and maybe one or two things caught your eye! I especially like the mapping trick for numpy labels and the multi-line trick for react native.
Comments