Available Knowhow
Spectral method results in efficient algorithm for faster results
|
Categories |
Computer vision, Artificial intelligence, Sensor and data systems |
|
Development Stage |
Proof of concept |
|
Knowhow |
Available Knowhow |
|
Market |
Algorithm has applications in financial forecasting, gene selection, sensor networks, resource allocation, machine learning and pattern recognition |
Highlights
-
Solution for the fast, efficient analysis of large quantities of data to identify the most relevant information for further analysis to provide specific information sought
-
Has been used in face recognition and analysis of gene expression data.
Our Innovation
Efficient, near-optimal greedy algorithm for sparse LDA analysis using a discrete spectral formulation based on variational eigenvalue bounds.
Key Features
-
Yields better than state-of-the-art performance on sparse PCA benchmarks used in the statistics community
-
Variational spectral method for finding optimal solutions using branch-and-bound search
-
Algorithm has been fruitfully applied to various real-world computer vision and machine learning tasks
The Opportunity
Applications in a wide variety of situations where relevant data must be identified from large amounts of inputs in order to provide a relevant analysis, such as stock market analysis, portfolio optimization, bankruptcy prediction, computer vision, face recognition, gene selection, analyzing consumer buying habits
1586128900.png)