Deep Learning Hands-on Workshop
summary
Hands-on setup hands out a pre-configured VirtualBox VM that boots Fedora and serves Jupyter notebooks with runnable Theano experiments; notebooks use Theano + Lasagne notebooks to load a 27 MB GoogleNet/Inception ImageNet model (ImageNet ~15M images, 1,000 classes) and train MNIST nets on 50,000 images (simple net ~91% rising to ~98% with conv layers). The workshop also builds a Bubble Breaker Q-learning agent trained on a GPU (~5 hours in the example) that raises mean scores from ~214 to ~384 and plays better than a casual human.
AI-written summary · about
chapters (12)
- 0:00Introduction to workshop
- 1:12Deep learning overview
- 2:29Applications of deep learning
- 7:57Neural network fundamentals
- 12:35Hands-on neural network demo
- 25:00Theano for computation
- 38:12Classifying digits with MNIST
- 45:07Convolutional neural networks
- 52:06ImageNet and deep learning models
- 1:07:18LSTMs and sequence processing
- 1:12:04Reinforcement learning in action
- 1:27:29Conclusion and Q&A
related talks
01h 07mDeep Learning - D.I.Y.
01h 30mFundamentals of Deep Learning Workshop
01h 29mHands-on introduction to Deep Learning with Keras and Tensorflow
02h 26mPractical Deep Learning for Data Scientists
01h 39mGoogle Colaboratory and Python: Deep Learning basics for CV tasks
01h 26mDeep Neural Networks with PyTorch
44mMachine Learning Exposed
01h 40mA Deep Dive into NLP with PyTorch