Machine Learning Test
Hi all this a machine learning test for grokvideo.ai. Please answer the questions by git commit-ing your answers separately. We would test your knowledge on
- Git
- machine learning tools
- problem solving (algorithmics)
These are mostly neglected skills that are not taught at schools. We would love to know your ability to take on learning or using them.
Git
- Write an algorithm to solve matrix multiplication using numpy
- Commit your result in a file named matmul.py in this directory
Machine Learning Tools
- Install Pytorch on your machine using
conda install pytorch==1.2.0 torchvision==0.4.0 -c pytorch
or follow the instructions here https://pytorch.org/get-started/previous-versions/ according to your OS.
- Write a logistic regression algorithm using Pytorch
- Write a 3 layered MLP using Pytorch to classify MNIST digits
- Try writing a dataloader that loads 256x256 images in a folder. Just give 2 images for example.
- Try doing number 4 but for Imagenet
Problem Solving
- Try doing matrix multiplication that doesn't use numpy and just arrays.
- What is the lower bound of matrix multiplication? (optional)
You should commit your files using the naming part_number.py
so for example you are solving Git for problem number 2 you name the file git_2.py
. We also would love to see a long git message.
Advanced Deep Learning (a plus)
- Write an autoencoder that takes a MFCC from an audio input (.flac or .mp3) and training it by minimizing the reconstruction loss?
- How can you reconstruct an audio from an MFCC? What algorithm is used?
- Write a Detectron algorithm for scene and object detection.
You should commit your files using the naming part_number.py
so for example you are solving Git for problem number 2 you name the file git_2.py
. We also would love to see a long git message.
Good luck! Bonne chance!