Deep Dive into Deep Learning with MXNet
summary
Hands-on demo of training and running models with Apache MXNet and the Deep Learning AMI on AWS, including launching GPU instances and an SSH tunnel; a multilayer perceptron and a convolutional network reach roughly 97% and 0.998 accuracy on MNIST in the notebook runs. A pretrained place-recognition model (YFCC100M / Multimedia Commons) predicts geographic coordinates from photos — demos hit landmarks like Hoover Dam though some inputs misclassify; deployment as a prediction service and integration with AWS services are shown.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 11:34What makes MXNet unique
- 21:10Deep learning applications
- 23:57How neural networks learn
- 35:28Training a model
- 43:46Demo with pre-trained models
- 1:06:36Image location prediction
- 1:13:21Building and tuning networks
- 1:15:15Q&A
related talks
18mModel Serving for Deep Learning
32mDeep Learning with TensorFlow, PyTorch and MXNet on AWS
29mAccelerate NumPy Data Science Workloads and Deep Learning Applications
23mBest Practices for Building Production-Grade Deep Learning Systems
42mAlex Smola, Professor in the Machine Learning Department, Carnegie Mellon University
54mMachine Learning for Developers
40mMachines are Learning
45mML: ML.NET: deep dive into dev AI