Microsoft DisruptAI Proof of Concept Project

Taken from Hardware for Machine Learning: Challenges and Opportunities^{^00} is a list of categories that are privacy, latency and security challenged:

1. APPLICATIONS

A. Computer Vision B. Speech Recognition C. Medical

2. MACHINE LEARNING BASICS

A. Feature Extraction B. Classification C. Deep Neural Networks (DNN) D. Complexity versus Difficulty of Task

3. OPPORTUNITIES IN ARCHITECTURES

A. CPU and GPU Platforms B. Accelerators

Use of regularization methods towards hidden signal processing sensor data^{^00} from image processing and EEG that can outperform the use of Diffusion Mapping (DM)

• use optimization/regularization methods
• signal processing data from image detection and EEG sentimental analysis records
• pinning hidden sensor data layer
• unsupervised deep learning
• use Locally Linear Embedding ALGM to outperform Diffusion Mapping

• ...

...

• user's emotions from EEG sentiment analysis
• hololens AR object classifications
• reinforcement learning

00: Source 00: arXiv:1612.07625v5 [cs.CV] 17 Oct 2017