Daily ML Tips @Daily_ML_Tips

Daily ML Tips from a Graduate student to help streamline your learning experience and spread knowledge without borders. Joined July 2019

Tweets

39
Followers

571
Following

159
Likes

27

Harrison Kinsley @Sentdex

6 years ago

LOL reddit.com/r/ProgrammerHu…

45 85 632 0 8

Download Image

Preethi Kasireddy @iam_preethi

6 years ago

If you don’t understand this, I am no longer your friend.

94 257 2K 0 38

Download Image

So in @sirajraval's livestream yesterday he mentioned his 'recent neural qubit paper'. I've found that huge chunks of it are plagiarised from a paper by Nathan Killoran, Seth Lloyd, and co-authors. E.g., in the attached images, red is Siraj, green is original

180 1K 4K 0 185

Download Image

Andrew M. Webb @AndrewM_Webb

6 years ago

@sirajraval Hey @esa, you're due to have this guy run a workshop soon. You should probably rethink that cosmos.esa.int/web/esac-stats…

7 27 426 0 2

Daily ML Tips @Daily_ML_Tips

6 years ago

Got a large dataset which takes many hours of training. A method I used to quickly get a prototype, is sampling the data and creating a model that over-fits. Then, once ready, i provide the rest of the data. Saves time and money. Thoughts? *take caution with domain

4 0 4 0 0

Daily ML Tips @Daily_ML_Tips

6 years ago

Why would you want to use dimensionality reduction techniques? Dimensionality reduction can allow you to: • Speed up training by reducing the number of features • Reduce memory usage • Identify latent features that impact multiple features

0 0 13 0 0

Daily ML Tips @Daily_ML_Tips

6 years ago

Who's tired of hyper parameter tuning? If you aren't, you're lying! Search methods are the best way to find params without putting in too much of the effort that comes with hyper parameter tuning. RandomSearchCV, GridSearchCV, GeneticSearchCV are some which can help you. :)

0 0 11 0 1

Daily ML Tips @Daily_ML_Tips

6 years ago

Not all tasks need neural networks, its surprising how many current business cases can be solved with other ML techniques as well as GROUPBY statements.

0 3 6 0 0

Daily ML Tips @Daily_ML_Tips

6 years ago

What new with Sci-Kit Learn? They finally have an API for plotting :') exciting times!

0 3 16 0 1

Download Image

Daily ML Tips @Daily_ML_Tips

6 years ago

Want to access all of NumPys functionality without importing it? With Pandas you can. This breaks the norm but is a fun trick ;)

1 0 14 0 0

Download Image

Daily ML Tips @Daily_ML_Tips

6 years ago

This is amazing.

arXiv Daily @Arxiv_Daily

6 years ago

This is amazing.

0 0 2 0 0

0 0 1 0 0

Daily ML Tips @Daily_ML_Tips

6 years ago

What projects are you guys working on right now? Trying any new techniques? :)

4 0 4 0 0

Daily ML Tips @Daily_ML_Tips

6 years ago

Unfortunately data sets with vague column names exist and dealing with them can get frustrating without proper documentation. Not anymore. By passing a Dict obj to the DataFrame.rename() function, you can tailor the data set as you like and make analysis fun. :)

1 1 18 0 1

Download Image

Daily ML Tips @Daily_ML_Tips

6 years ago

Something different. People have DM'd asking for some general python tips too. There is a misconception that to be good at coding you have to know all functions. IMO, you should know how to apply functions no matter the language, use dir() or help() to look up functions.

0 1 8 0 0

Download Image

Daily ML Tips @Daily_ML_Tips

6 years ago

Ever wanted to quickly analyze data you saw online or somewhere not as accessible. pandas offers the read_clipboard() functions which allows you to make a data frame with ease. From the screenshot you'll see I had data in Excel which I copied (Ctrl + C). :)

3 4 25 0 2

Download Image

Daily ML Tips @Daily_ML_Tips

6 years ago

Following on from the previous memory management post, Chunks. By passing the argument chunksize we can split our input data into chunks to avoid loading a large file into memory. Note the change in the type() of both our data frames and the shape of our data frames. Enjoy :)