SBIR-STTR Award

We propose a principled in situ learning framework that is appropriate for a Bayesian classifier implemented with semi-supervised and multi-task learning. We will investigate several different forms of in situ learning, and will perform testing on measured data to help define which is most appropriate for Navy sensing missions. In addition, we will develop new techniques for feature adaptivity and selection, to tune the features to the particular targets and clutter in the environment under test.

Benefit:
The proposed research has the opportunity to significantly advance the manner in which the Navy performs underwater sensing. The proposed algorithms will yield a high level of adaptivity, allowing for changing environmental, target and clutter conditions. This will be performed using a new class of in situ learning algorithms, that allow the classifiers to adapt to sensing conditions. The proposed algorithms are appropriate for integration within semi-supervised and multi-task learning, these exploiting a significant level of context. In the proposed research we will not explicitly develop semi-supervised and multi-task algorithms, as these are being developed by SIG under separate support. However, we will leverage that research, and the proposed in situ learning algorithms will be integrated into such. The proposed Phase I research will make the following specific contributions: (1) The proposed algorithms will take into account real-world sensing issues, such as the potential for noisy data/labels, as well as the cost of label/data acquisition. (2) A new Bayesian Elastic Net algorithm is proposed, which will provide a level of accuracy in feature selection that has previously been unavailable. In a Phase II effort, these new approaches will be integrated into the semi-supervised multi-task learning algorithms, providing a unique framework for sensing adaptivity and in situ learning. The potential commercial applications of the research are as follows. The availability of high spatial and spectral resolution satellite imagery has created a growth industry in land-use assessment, optimized natural resource extraction, habitat analysis, precision agriculture, and urban planning/infrastructure analysis. However, the variability of sensing parameters (i.e., cloud-cover, seasonal variations, incident and scattering angles relative to time of day and sensor orientation) can confound classification performance between training vs. operational imagery. Moreover, the amount of labeled training data is small relative to the volume of spatial/spectral imagery generated by modern commercial satellites such as Ikonosand LandSat 7. The algorithms developed under this effort could accommodate such high data volumes and adapt to sensor variability for improved image scene characterization. Such methods could be useful to the Bureau of Land management (BLM), the United States Geologic Service (USGS), the United States Department of Agriculture (USDA), and the intelligence community for the analysis of commercial, and non-commercial, remote sensing data. Another key application of the proposed research is in the area of bioinformatics. For example, when dealing with gene-expression data one typically has large quantities of highdimensional gene-expression feature vectors, with few labeled training examples. The labeling of this data is very expensive and labor intensive, requiring the attention of experts. The in situ learning algorithms may be used to provide feedback to medical researchers, as to which feature vectors are most informative if they could be labeled (for example, by a human expert, or via conventional diagnoses). Moreover, the large quantities of unlabeled data make this a good application for semisupervised algorithms, in which all available data are exploited when performing algorithm design (both labeled and unlabeled data). Initial forms of these algorithms were developed recently by Prof. Carin at Duke, and several of his recent PhD graduates have been employed by medical companies such as SIEMENS, GE and Guidant. It is therefore anticipated that such companies represent a wide market for the class of software tools we propose to develop under this effort.

Keywords:
adaptive, adaptive, learning, Acoustic, Bayesian

In the proposed Phase II program, the methods developed and implemented during Phase I research will be fully integrated within a common Bayesian in situ learning framework. We have developed several Bayesian classifiers, to which we will apply label acquisition and label confidence techniques. Additionally, we will extend the in situ learning framework to include multi-task learning. Previously collected sensing data are often available from different sensors or environments. Not all data are related, however the potential exists to share information between related tasks and exploit the contextual information of previous tasks. The current in situ learning process is inherently myopic; the algorithm identifies the single most-informative data sample. The ability to select multiple samples without relearning the classifier can increase computational efficiency and maximize analyst workload. Based on the theory of submodular functions, non-myopic in situ learning techniques for subset selection will be developed and integrated into the Bayesian framework. Finally, new statistical embedding technology will be investigated that allows an analyst to synthesize data for training and to augment the label acquisition process. A low-dimensional embedded space may be visualized, and any location on the manifold can be recreated in the original high-dimensional space.

Benefit:
The in situ learning methods may provide significant performance improvements for automated detection and classification of anomalous tissues in medical imagery. SIG has been in discussions with Research Triangle Park University researchers and private consultants to investigate new methods for feature extraction and sensor tuning for new coherent x-ray mammography techniques. Active learning methods can provide a mechanism to associate imagery acquired from different aspects and to capture the inherent spatial and textural dependencies of such imagery to provide improved cancer detection performance. Active learning methods can provide a means to explicitly incorporate radiologist feedback into the detection and classification design process. Potential customers of such medical imaging products include GE medical systems, Siemens, and collateral markets in pharmaceutical development. SIG has ties to technical staff within both GE and Siemens, and Research Triangle Park, NC serves as a hub to several major pharmaceutical companies. Moreover, SIG has close ties with the NC venture capital (VC) communities that fund both information technology and life sciences innovations. As the technology developed in all Phases of this SBIR research evolves, SIG will discuss relevant advances with the medical imaging and VC communities to identify further commercial opportunities. SIG also has extensive experience in wide-area motion imagery, and target tracking and identification. SIG has had discussions with IBM and Government Technology Solutions, Inc. (GTSI) to investigate the potential to transition its WAPS technologies to large-scale municipal public safety systems in North America and Europe. Many of these technologies involve data from multiple sensors and the potential to ingest input from a human operator to achieve specific mission objectives. There is also a current market for target advertising by automatically mining video content, including online video (e.g. YouTube) and network television. There are several potential new markets for automated airborne video exploitation, including traffic analysis and modeling for traffic flow optimization and investigation of locations for new store locations/developments, and automated location and tracking of dismounts and vehicles for FBI, sheriff, and municipal police force applications. SIG has an ongoing relationship with companies interested in participating in these potential new markets. Further definition of market and product requirements with traffic analysis decision makers will be pursued to better assess these transition opportunities.

Keywords:
semi-supervised learning, elastic net, active learning, submodularity, statistical embedding, hierarchical Bayesian models, multi-task learning