Deep Learning Measures of Effectiveness

The resurgence of interest in artificial intelligence (AI) stem from impressive deep learning (DL) performance such as hierarchical supervised training using a Convolutional Neural Network (CNN). Current DL needs to focus on contextual reasoning, explainable results, and repeatable understanding that require evaluation methods. This paper presents measures of effectiveness (MOE) for DL techniques that extend measures of performance (MOP). MOPs include: Timeliness: computational efficiency, Accuracy: operational robustness, and Confidence: semi-supervised representation. MOE concerns include Throughput: data efficiency, Security: adversarial robustness, and Completeness: problem representation. DL evaluation requires verification and validation testing in realistic environments. An example is shown for Deep Multimodal Image Fusion (DMIF) that evaluates MOEs of information gain, robustness, and quality.