A Keras usage pattern that allows for maximum flexibility when defining arbitrary losses and metrics (that don't match the usual signature) is the "endpoint layer" pattern. It works like this: https://t.co/dhYFKeemnC pic.twitter.com/v3YtMrrcQb
— François Chollet (@fchollet) May 14, 2019
To get more from `itertools` library, see its 'recipes' section, https://t.co/L5guQ06MT9
— Daily Python Tip (@python_tip) May 14, 2019
and `repeatfunc` as an example.
https://t.co/2JRyh1yNsj
my latest blog post: Advice for Better Blog Postshttps://t.co/shIlmrw4Y8
— Rachel Thomas (@math_rachel) May 13, 2019
You can select elements of 2d #numpy array by lists of row and column indices
— Daily Python Tip (@python_tip) May 13, 2019
>>> x = np.arange(9).reshape((3, 3))
>>> x
array([[0, 1, 2],
[3, 4, 5],
[6, 7, 8]])
>>> x[[0,2,2], [1,0,2]]
array([1, 6, 8])
For details: https://t.co/LKY2rj9l8l
Nice tip from the replies - janitor::clean_names() does this and more! If you haven't seen janitor by Sam Firke, run, do not walk - it also has functions to remove empty rows and columns, find duplicates, and create frequency tables.https://t.co/TPJ6UrfKk8 https://t.co/cAtcCjydFn
— Emily Robinson (@robinson_es) April 30, 2019
One line change: pic.twitter.com/2DWLuQcjue
— Delip Rao (@deliprao) April 29, 2019
🎨 A short thread of some cool ways to visualize @spacy_io's output in your terminal! pic.twitter.com/M7naCs2wPJ
— Ines Montani 〰️ (@_inesmontani) April 29, 2019
So much of being a successful employee revolves around effective communication. If you're rude to someone you're trying to get to hire you, how are you going to treat colleagues who you may have power over?
— jacobian (@jacobian) April 29, 2019
New blog post: "A Recipe for Training Neural Networks" https://t.co/5lBy4J77aS a collection of attempted advice for training neural nets with a focus on how to structure that process over time
— Andrej Karpathy (@karpathy) April 25, 2019
Tricky Python bug I just hit due to an incorrect mental model of class & instance attributes:
— Jake VanderPlas (@jakevdp) April 24, 2019
class Foo:
_num_instances = 0
def __init__(self):
self._num_instances += 1
f1 = Foo()
f2 = Foo()
print(Foo._num_instances)
Sometimes even 5 rows is too many to show as the default for the DataFrame head method. Use partialmethod from functools to change this.
— Ted Petrou (@TedPetrou) April 19, 2019
pd.DataFrame.head = partialmethod(pd.DataFrame.head, n=3) pic.twitter.com/NPKf9Jlg5F
Totally agree. Both because it makes tactical sense, and because this is how you should do it as a nice human being. https://t.co/ZkP1xzBhiM
— Julian Togelius (@togelius) April 17, 2019
