I see a max_error parameter in the kmeans function definition set to a default value of 0.01 which the candidate doesn’t use or talk about. What is that parameter?
Nice mock! However, at one point in video the interviewee said that we can use any distance in distance_function but what if I choose Cosine-distance? How would you calculate centroids? Choosing Cosine-distance would change this to a spherical clustering problem and the algorithm would change. So it's better to be sure that we know what we say in an interview .
You can choose cosine distance when dealing with text data in the form of word embeddings. Calculate of centroids is still the mean point across all dimensions
Don't leave your machine learning career to chance. Sign up for Exponent's Data Science interview course today: bit.ly/3iMOoPQ
The interview looks authentic. Thank you!
Not it doesn't, if you are some senior engineer, you won't pick centroids randomly.
I see a max_error parameter in the kmeans function definition set to a default value of 0.01 which the candidate doesn’t use or talk about. What is that parameter?
Nice mock!
However, at one point in video the interviewee said that we can use any distance in distance_function but what if I choose Cosine-distance? How would you calculate centroids?
Choosing Cosine-distance would change this to a spherical clustering problem and the algorithm would change. So it's better to be sure that we know what we say in an interview .
You can choose cosine distance when dealing with text data in the form of word embeddings. Calculate of centroids is still the mean point across all dimensions
Actually, you cannot choose cosine distance. K-means works only with euclidean distance. This is what "k-means" means.
Yup something about not being able to converge if using alternative distance function
🚀Love it!
I see very little point to coding k-means.
absolutely no point. Such screenings are useless...
@@Squeed79 Do you think such will be asked for ML engineer roles????
@@akhilpadmanaban3242 for ML engineer? I would not ask "coding" from scratch. Maybe only principles and if/how it is scalable to huge data, etc...
points[points_i], passing list to a list or what, I mn sure it works here, and I am missing something but I don't know what please help