Se ha denunciado esta presentación.
Utilizamos tu perfil de LinkedIn y tus datos de actividad para personalizar los anuncios y mostrarte publicidad más relevante. Puedes cambiar tus preferencias de publicidad en cualquier momento.

First 5 years of PSI:ML - Filip Panjevic

Filip Panjevic is a Co-Founder and CTO at ydrive.ai - startup dealing with self-driving cars, and one of the founders of Petnica Machine Learning School.

Filip's talk will focus on the story of Petnica School, how did it start, what has changed since the beginning, how the concept of school looks right now and why is that concept good for making new data scientists. This talk will be perfect for people who consider starting their careers in the data science field!

  • Sé el primero en comentar

First 5 years of PSI:ML - Filip Panjevic

  1. 1. Filip Panjevic First 5 years of PSI:ML DATA SCIENCE /CONFERENCE/
  2. 2. 𝛙 : Machine Learning
  3. 3. 𝛙 : Machine Learning 10 day machine learning summer school
  4. 4. 𝛙 : Machine Learning 10 day machine learning summer school Theory and hands-on experience with mentorship
  5. 5. 𝛙 : Machine Learning 10 day machine learning summer school Theory and hands-on experience with mentorship Teachers from Microsoft, MSR, Belgrade U, Cambridge U, DeepMind, Google Brain, etc.
  6. 6. 𝛙 : Machine Learning 10 day machine learning summer school Theory and hands-on experience with mentorship Teachers from Microsoft, MSR, Belgrade U, Cambridge U, DeepMind, Google Brain, etc. 300+ applicants each year from all over the world
  7. 7. 𝛙 : Machine Learning 10 day machine learning summer school Theory and hands-on experience with mentorship Teachers from Microsoft, MSR, Belgrade U, Cambridge U, DeepMind, Google Brain, etc. 300+ applicants each year from all over the world 130+ graduates to date
  8. 8. 𝛙 : Machine Learning 10 day machine learning summer school Theory and hands-on experience with mentorship Teachers from Microsoft, MSR, Belgrade U, Cambridge U, DeepMind, Google Brain, etc. 300+ applicants each year from all over the world 130+ graduates to date 91% of alumni believe PSI:ML influenced their education and their careers
  9. 9. Motivation
  10. 10. Year One
  11. 11. Lectures and workshops Projects Teachers and mentors Venue Funding Dates Advertising Branding Application process Selection process =
  12. 12. Lectures and workshops Projects Teachers and mentors Venue Funding Dates Advertising Branding Application process Selection process =
  13. 13. Microsoft Petnica
  14. 14. Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9 Day 10
  15. 15. Lectures and workshops Projects Teachers and mentors Venue Funding Dates Advertising Branding Application process Selection process =
  16. 16. Lectures and workshops Projects Teachers and mentors Venue Funding Dates Advertising Branding Application process Selection process =
  17. 17. Lectures and workshops Projects Teachers and mentors Venue Funding Dates Advertising Branding Application process Selection process =
  18. 18. Lectures and workshops Projects Teachers and mentors Venue Funding Dates Advertising Branding Application process Selection process =
  19. 19. Ideal candidate Undergrad or grad student Interested in ML Good with math Likes to code/hack Thinks outside the box Can absorb a densely packed curriculum Works well in a team
  20. 20. Ideal candidate Undergrad or grad student Interested in ML Good with math Likes to code/hack Thinks outside the box Can absorb a densely packed curriculum Works well in a team
  21. 21. White background BMP image Dark square of unknown size
  22. 22. White background BMP image Tic-tac-toe board Determine who won the round
  23. 23. 2015 Student Projects IMDB comment sentiment classification Website language detection Math symbol recognizer Whiteboard background removal Gesture controlled “Flappy Bird”
  24. 24. Feedback
  25. 25. “Pairing lectures with workshops so it is easier to grasp the theoretical framework with hands on approach” “Workshops had limited interactivity, mostly came down to copy-pasting” “Schedule too crammed - should be at least 3-4 more days” “Too diverse lectures - should be split into basic and intermediate” “We might consider two sessions e.g. beginners and advanced.” “Projects should have started earlier so they had more time to understand the problem and better organize.”
  26. 26. Iterating
  27. 27. Introduction to ML Introduction to Data Science Logistic Regression Neural Networks Convolutional Neural Networks Recurrent Neural Networks Random Decision Forests Clustering Principal Component Analysis & Autoencoders Support Vector Machines Logistic regression workshop Caffe workshop Curriculum 2015
  28. 28. Introduction to ML Introduction to Data Science Logistic Regression Neural Networks Convolutional Neural Networks Recurrent Neural Networks Random Decision Forests Clustering Principal Component Analysis & Autoencoders Support Vector Machines Bayesian data analysis Backpropagation Markov Chain Monte Carlo Natural Language Processing Supervised Learning Algorithms Logistic regression workshop Caffe workshop Octave workshop RNN workshop Curriculum 2016
  29. 29. Introduction to ML Logistic Regression Neural Networks Convolutional Neural Networks Recurrent Neural Networks Semantic Segmentation Object Detection Generative Adversarial Networks Random Decision Forests Clustering & Principal Component Analysis Support Vector Machines Bayesian data analysis Backpropagation Markov Chain Monte Carlo Natural Language Processing Supervised Learning Algorithms Reinforcement Learning Logistic regression workshop Clustering & PCA workshop Caffe workshop Octave workshop RNN workshop NN workshop GAN workshop Curriculum 2017
  30. 30. Introduction to ML Logistic Regression Neural Networks Convolutional Neural Networks Recurrent Neural Networks Semantic Segmentation Object Detection Generative Adversarial Networks Random Decision Forests Clustering & Principal Component Analysis Support Vector Machines Natural Language Processing Supervised Learning Algorithms Reinforcement Learning SLAM Machine Learning in Medicine Logistic regression workshop Clustering & PCA workshop NN workshop RNN workshop GAN workshop CNN workshop RL workshop NLP workshop Curriculum 2018
  31. 31. Introduction to ML Logistic Regression Neural Networks Convolutional Neural Networks Recurrent Neural Networks Generative Adversarial Networks Random Decision Forests Boosting Gaussian Processes Clustering & Principal Component Analysis Support Vector Machines Modern Natural Language Processing Supervised Learning Algorithms Reinforcement Learning Geometric ML Machine Learning in Medicine Numpy Intro TensorFlow intro Logistic regression workshop NN workshop RNN workshop GAN workshop CNN workshop RL workshop NLP workshop Curriculum 2019
  32. 32. Homework 2015 1x Outside-of-the-box problem Solutions sent as *.exe files over email Renamed to pass spam filters Tested semi-manually
  33. 33. Homework 2016 2x Outside-of-the-box problem 1x Pure engineering problem Solutions submitted over email in Python, C, C++, Octave... Tested semi-manually
  34. 34. Homework 2017 2x Outside-of-the-box problem 1x Pure engineering problem Solutions submitted through Petlja.org in Python, C, C++, Octave... Tested automatically through Petlja.org
  35. 35. Homework 2018 2x Outside-of-the-box problem 1x Pure engineering problem Solutions submitted through Petlja.org in Python, C, C++, Octave... Tested automatically through Petlja.org
  36. 36. Homework 2019 2x Outside-of-the-box problem 1x Pure engineering problem Solutions submitted through Petlja.org in Python Tested automatically through Petlja.org Tasks from previous years available on Petlja.org
  37. 37. +
  38. 38. +
  39. 39. 2019 Student Projects Solving Rubik’s cube using RL Learning to walk using RL Generating Favicons using GAN’s Generating Anime characters using GAN’s Depth estimation from stereo Vehicle egomotion estimation
  40. 40. “I went from not being accepted to PSI:ML to actually becoming a lecturer there! Career wise – applying for the Summer Institute turned out to be one of the best decisions I've made – it jump-started me into neural networks.” -Bruno Gavranovic
  41. 41. “There’s a lot of content to go through and always something to discuss, but don’t fret, it all starts from the very basics of ML. So, even if you don’t have any idea about anything, the seminar will pick you up at whatever skill level you’re at.” -Vladimir Nikolić and Leander Schröder
  42. 42. “Now I am a part of the Evoke team in Microsoft Development Center Serbia, where I am seeing first hand how ML is used every day for improving products and user experience.” -Natalija Radić
  43. 43. “The most important part of the seminar is the people. There, you have a great opportunity to grow your professional network and make new friends by meeting a handful of hardworking and talented participants and lecturers.” -Marko Mihajlović and Nikola Popović
  44. 44. PSI:ML 2046
  45. 45. DATA SCIENCE /CONFERENCE/ THANK YOU

×