H2O Deep Water with Python early sneek
summary
Python code defines networks that run remotely on an H2O cluster using TensorFlow MXNet or Caffe; the demo loads pre‑built architectures (LeNet AlexNet VGG Inception ResNet V2), trains on CPU or GPU, and reports ~80% top‑1 and 95% top‑5 for modern nets versus ~94% human. Shows an autoencoder on Olivetti faces that compresses and reconstructs images, remote grid and random hyperparameter search with checkpointing, and transfer‑learning by importing pretrained weights into Deep Water.
AI-written summary · about
chapters (19)
- 0:00Deep learning excitement
- 1:06Why deep learning wins
- 2:51Neural network capacity
- 5:06CNNs for image processing
- 7:38Automatic gradient computation
- 9:02Image classification breakthroughs
- 10:51Self-learning features
- 13:41Autoencoders and compression
- 16:10H2O Deep Water autoencoders demo
- 19:40Deploying models in production
- 22:09Deep Water framework
- 24:00CIFAR-10 with H2O Deep Water
- 27:39Transfer learning
- 28:43Building networks with MXNet
- 31:12Hyperparameter tuning
- 34:26Random grid search
- 35:48Model checkpointing
- 36:33Open-source resources
- 36:55Q&A
related talks
42mDeep Learning is Like Water (it's Everywhere!)
50mScalable Data Science & Deep Learning with H2O
01h 18mIntroduction to H20 AutoML with Python
01h 10mIntroduction to Machine Learning with H2O and Python
45mMassive distributed processing with H2O
01h 26mMachine Learning Algorithms with H2o and PowerShell
44mScalable Automatic Machine Learning with H2O
30mScalable Automatic Machine Learning with H2O