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:
- APPLICATIONS
A. Computer Vision B. Speech Recognition C. Medical
- MACHINE LEARNING BASICS
A. Feature Extraction B. Classification C. Deep Neural Networks (DNN) D. Complexity versus Difficulty of Task
- 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