- Let's Learn about Self Driving Cars
I have recently finished the Udacity Self Driving Car Nano degree (https://www.udacity.com/course/self-driving-car-engineer-nanodegree--nd013). I would like to share what I learned there. You might be interested to know that both Coursera and Edx are offering self-driving car courses.
- AI tools and other topics
We will examine cloud tools and other topics. Agenda: 1- We will explore tools to run Deep Learning algorithms in the cloud. Among them AWS SageMaker (paid) and Google Colab (free). 2- We will explore for those who have GPUS machines how to run nvidia-docker PS: - This session is meant to be a learning experience for all of us. So we will workshop and explore together. - We could cover how to set up an amazon account and a google account, but I rather people come with those acounts set up. - Nvidia based laptop is not a requisite, if people don't have a GPU, we can skip this. - There is internet at the cafe - Please RSVP only if you intend to attend. Montreal - Let's learn AI
- Classical Computer Vision + lessons learned from DeepLearning.ai
We will discuss classic computer vision algorithms implemented in these two videos: https://youtu.be/WzbgZtzRC1k https://youtu.be/jo9zpY3WC20 Essentially classic computer vision, feature engineering and SVMs. Why are they effective and what are their real shortcomings. We will also discuss lessons learned from Andrew Ng's Deep Learning AI series of courses on Coursera. Edit: 1- Guys on the waiting list, if it's just two people I suggest you attend, we will try to squeeze into the room (which is pretty big) 2- Git repos: here's the udacity starter kit for term 1: https://github.com/udacity/CarND-Term1-Starter-Kit here are the two project git repos: https://github.com/udacity/CarND-LaneLines-P1 https://github.com/udacity/CarND-Vehicle-Detection my solutions: https://github.com/lets-learn-AI/advanced-lane-detection https://github.com/lets-learn-AI/vehicle-detection-svm here's Udacity image detection done right using Yolo: https://github.com/udacity/self-driving-car/tree/master/vehicle-detection/darkflow (note that Yolo was covered in one of the courses by Andrew Ng and we will discuss that) Edit 2: - Guys the meetup is what we make of it. If people come prepared, it will be a workshop session around getting the first classical CV and ML projects working. - If anyone can't attend please change your RSVP status.
- Self Driving Car anyone?
This is the first meetup about the self-driving car. We will cover an exercise where we have implemented a convolutional neural network that drives a simulation of a car. We will also have an open discussion about self-driving car technologies and challenges. Come to have a good time and learn. Montreal Let's Learn AI... C'est le premier meetup sur la voiture autonome. Nous couvrirons un exercice où nous avons mis en œuvre un réseau de neurones convolutionnels qui conduit une simulation d'une voiture. Nous aurons également une discussion ouverte sur les technologies et les défis liés aux voitures autonomes. Venez passer un bon moment et apprendre. Montréal Apprenons AI ... Edit: 0- environment setup MUST DO PREREQUISITEhttps://github.com/udacity/CarND-Term1-Starter-Kit 1- My solution to the behavioral cloning exercise: https://github.com/lets-learn-AI/behavioural-cloning.git 2- udacity github for the project: https://github.com/udacity/CarND-Behavioral-Cloning-P3 3- simulator from github: https://github.com/udacity/self-driving-car-sim (I can't privide the exe as I'm not sure of the license, you need to build this...) 4- Nvidia architecture that udacity suggests to implement to solve the problem: https://devblogs.nvidia.com/deep-learning-self-driving-cars/ PS: - I suggest to come prepared. to have the capacity to access the internet, either by getting a library card or have enough data on the cell phone and to have a PC that can run the simulator and the neural network, you might need a GPU (i did) Finally here's a video of the result I got: https://youtu.be/Qb2RntKnl4M
- So you think you know backpropagation?
Hi everyone, For our first Meetup of the year, we try a new recipe: a talk in the big room! The talk is about *backpropagation* in general and our main goal is that every attendee leaves with a strong understanding of this algorithm, at least in its basic form. This talk is not only about backpropagation but its combination with gradient descent methods to minimize a loss function in a neural network. We will detail and visualize all these terms in great detail. The talk is divided into two main parts: - Basic ideas: this is an introduction to backprogation. - Some variations: after understanding the basics, we elaborate a little bit more on the subject. There is a third part called "Are you nuts?" in case we run out of material (this is unlikely but could serve as material for a new talk if some are interested) and even a fourth hidden part! Yep! Notice that the Meetup is happening on a WEDNESDAY (not Tuesday)! Material: - TO COME Montreal Let's Learn AI.
- RNN again: many to many
Hi everyone, the room we have will fit 25. We will be up on the 5th floor. Notice that the meetup is happening on a TUESDAY (not Wednesday)! Material: - Demo on how to use Floydhub for quickly training examples. - Work through a tutorial for sequence-to-sequence from Udacity. - Comparison with Keras implementation of similar network This will build upon our last meetups on RNNs and LSTMs and progress towards solving practical problems with RNNs. Code and resources can be found on our github repository: https://github.com/Mtl-lets-learn-AI/2017_11_28_seq2seq Pretrained weights for the Udacity code can be found here. https://www.floydhub.com/nma38/projects/mllai-seq2seq/1/output Montreal Let's Learn AI.
- Intuition for Long Short Term Memory (LSTM)
Hi everyone, the room we have will fit 20. We will be up on the 5th floor. Since space is limited and we would like your RVSPs in a day before the event! We will meet and share our understandings of LSTMs and work through a Udacity tutorial on Sentiment Analysis using it. https://github.com/udacity/deep-learning/tree/master/sentiment-rnn Recommended Reading: http://colah.github.io/posts/2015-08-Understanding-LSTMs/ This will build upon our last meetup on RNNs and progress towards solving practical problems with RNNs. Code and resources can be found on our github repository: Montreal Let's Learn AI.
- Recurrent Neural Networks (RNNs)
Hi everyone we have a new location at the Microsoft Montreal office! Please be advised that space is limited and we would like your RVSPs in a day before the event! We will meet and work through a tutorial to understand RNNs. “How to build a Recurrent Neural Network in TensorFlow (1/7)” https://medium.com/@erikhallstrm/hello-world-rnn-83cd7105b767 “Using the RNN API in TensorFlow (2/7)” https://medium.com/@erikhallstrm/tensorflow-rnn-api-2bb31821b185 This will build upon our last meetup on Wordembeddings and progress towards understanding LSTMs. Code and resources can be found on our github repository: https://github.com/Mtl-lets-learn-AI/2017_11_01_RNN . Montreal Let's Learn AI.
- Word embeding
We will meet and look at this tutorial https://github.com/udacity/deep-learning/tree/master/embeddings that explain the skip-gram algorithm for word vector. Try to install everything needed to run the notebook (the one with the solutions). After, we will look the utilisation of the gensim library that simplify a lot the creation of vector for word. So try also to have/install the libraries (before the meetup) : gensim, numpy, pandas, scikit-learn, maplotlib and mpld3. We will be at the room[masked] from 19:00 to 21:00. Montreal Let's Learn AI.