Areas of focus:
- Natural Language Processing
- Generative Adversarial Networks
- Recommender Systems
- Reinforcement Learning
Since this repository contains submodules use the following command to clone:
git clone --recurse-submodules https://github.com/Ishitori/MXNetWorkshopHongKong.git
Area of focus: Get introduced to MXNet (Gluon API) on AWS using Amazon SageMaker. Presenters: Thom, Thomas, Cyrus, Sergey
| Title | Description | Duration (Mins) | Presenter |
|---|---|---|---|
| MXNet & Gluon Overview | An overview of MXNet architecture and components | 30 | Thom |
| SageMaker Setup | Get configured with SageMaker for the day ahead | 30 | Cyrus |
| Gluon Crash Course | Walkthrough of core Gluon components, and use them to create and train a convolutional neural network. | 180 | Thomas |
| SageMaker Introduction | An introduction to SageMaker SDK | 30 | Cyrus |
| Multi GPU training | An introduction to training using multiple GPUs in Gluon, with lab. | 60 | Sergey |
| Multi GPU training with SageMaker | Same as above on SageMaker | 30 | Sergey |
Area of focus: Natural Language Processing examples Presenters: Sergey, Thom and Cyrus
| Title | Description | Duration (Mins) | Presenter |
|---|---|---|---|
| Deep Learning AMI Setup | An introduction to DLAMI and get configured for the day ahead | 30 | Thom |
| PyCharm Setup | Setup for PyCharm for Remote Debugging | 30 | Thom |
| Gluon NLP | Introduction into solving common NLP tasks using GluonNLP library | 30 | Sergey & Cyrus |
| Stacked Bidirectional LSTM (LAB) | We work up from plain RNN to Stacked Bi-directional LSTMs using Gluon layers | 60 | Thom |
| Keyphrase Extraction | Will show how to implement a model from Bi-directional LSTM recurrent neural network for kyphrase extraction paper, using Gluon components and cover how to do common NLP tasks like Data pipelining, Tokenization, Embedding, modeling using LSTM. | 60 | Sergey |
| MXBoard | Using MXNet with Tensorboard to monitor training | 30 | Thom |
| Sentiment Analysis in code-switching text | We will take part in NLPCC 2018 Emotion Detection in Code-Switching Text competition, dive deep into working with Embeddings, and learn how to use Convolutional encoder from this paper | 60 | Sergey |
| Beyond the Defaults | Alternatives for initialization, optimization and evaluation metrics. | 30 | Cyrus |
| Intent detection and slot filling | We will see how to use advanced embedding like ELMo to transfer learning in NLP, do multitasking and use Highway layer and Conditional Random Field. | 60 | Sergey |
Area of focus: Recommender Systems Presenters: Cyrus, Thom and Sergey
| Title | Description | Duration (Mins) | Presenter |
|---|---|---|---|
| Survey of Recomender Systems | 60 | Cyrus | |
| Profiling MXNet | Analysis of runtime code to identify performance bottlenecks | 30 | Thom |
| Implementation of MLP on Movie Lens | 30 | Cyrus | |
| MXNet Performance Tricks | Useful tips for maximizing performance of MXNet | 60 | Sergey |
| DSSM Theory | 30 | Cyrus | |
| LR Schedules | Using custom learning rate schdules | 45 | Thom |
| DSSM Implementation | 60 | Cyrus | |
| Sparse Matrix Operations | 45 | Cyrus |
Area of focus: Generative Adversarial Networks, Model Deployment and multi-host training. Presenters: Thom, Sergey & Steve
| Title | Description | Duration (Mins) | Presenter |
|---|---|---|---|
| InfoGAN Theory | A paper walkthrough of InfoGAN | 30 | Thom |
| InfoGAN Implementation | A code walkthrough of Gluon implementation of InfoGAN | 60 | Thom |
| ECommerce GAN Theory | A paper walkthrough of e-Commerce GAN | 30 | Sergey |
| SageMaker Automatic Model Tuning | A look at SageMaker's hyperparameter optimization features. | 60 | Thom |
| Multi Host training | Using multiple instances to speed up training (on SageMaker) | 60 | Sergey |
| SageMaker Deployment | Useful patterns for deployment on SageMaker. | 30 | Steve |
| MMS Deployment | MXNet Model Server deployment on AWS Fargate | 60 | Sergey |
Area of focus: Reinforcement Learning Presenters: Thom, Sergey & Steve
| Title | Description | Duration (Mins) | Presenter |
|---|---|---|---|
| re:Invent ML Announcements | Recap of new ML features | 60 | Steve |
| AWS Neo & TVM | An introduction to model compiliation with AWS Neo | 30 | Thom |
| AWS Elastic Inference | 30 | Steve | |
| RL Introduction | An introduction to RL theory, and SageMaker RL (frameworks) | 45 | Thom |
| RL DQN | An introduction to DQN with SageMaker RL Coach demo. | 45 | Thom |
| RL PPO | An introduction to PPO with SageMaker RL Coach demo. | 45 | Thom |