Much of the previous machine learning (ML) fairness literature assumes that protected features such as race and sex are present in the dataset, and relies upon them to mitigate fairness concerns. However, in practice factors like privacy and regulation often preclude the collection of protected features, or their use for training or inference, severely limiting the applicability of traditional fairness research. Therefore, we ask: How can we train a ML model to improve fairness when we do not even know the protected group memberships? In this work we address this problem by proposing Adversarially Reweighted Learning (ARL). In particular, we hypothesize that non-protected features and task labels are valuable for identifying fairness issues, and can be used to co-train an adversarial reweighting approach for improving fairness. Our results show that ARL improves Rawlsian Max-Min fairness, with notable AUC improvements for worst-case protected groups in multiple datasets, outperforming state-of-the-art alternatives.

Factory and process automation systems are increasingly employing information and communications technologies to facilitate data sharing and analysis in integrated control operations. Wireless connections provide flexible access to a variety of field instruments and reduce network installation and maintenance costs. This serves as an incentive for the adoption of industrial wireless networks based on standards such as the WirelessHART and ISA100.11a in factory control systems. However, process control systems vary greatly and have diverse wireless networking requirements in different applications. These requirements include deterministic transmissions in the shared wireless bandwidth, low-cost operation, long-term durability, and high reliability in the harsh radio propagation environment. It is an open question whether a generic wireless technology would meet the requirements of industrial process control. In this paper, we propose a novel simulation framework for performance evaluation of wireless networks in factory and process automation systems. We select a typical process control plant model, specifically the Tennessee Eastman Challenge (TE) Model, and define the interfaces between the process simulator and the wireless network simulator. We develop a model of the protocol stack of the WirelessHART specification in the OMNET++ simulation engine as a typical industrial wireless network. We present simulation results that validate the prospect of using WirelessHART in the TE plant, and we evaluate the impact of various wireless network configurations on the plant operation. Given its modular design, the proposed simulation framework can be easily used to evaluate the performance of other industrial wireless networks in conjunction with a variety of process control systems.

The classification of texture images, especially those with spatial rotation and region shift, is a challenge and important problem in image analysis and classification. This paper proposes a novel algorithm design, an ellipse invariant algorithm, to improve the capability of texture classification for spatial rotation and region shift. The principle of an ellipse invariant algorithm is to use a minimum ellipse to enclose specific representative pixels extracted by the subtracting clustering method. After translating the coordinates, the ellipse in the rotated texture would be formulated as the ellipse in original texture. Also in this paper a hybrid texture filter is proposed. In the hybrid texture filter the scheme of texture feature extraction include Gabor wavelet, neighboring grey level dependence matrix and the ellipse invariant algorithm. Support vector machines (SVMs) are introduced as the classifier. The proposed hybrid texture filter can classify both the stochastic textures and structural textures. Experimental results reveal that this proposed algorithm outperforms existing design algorithms.

The architecture of a Software Defined Radio (SDR) multi-standard baseband processor are presented. As the first representative of a new SDR baseband family, the X-GOLD™ SDR20 has been successfully designed and fabricated in a 65nm CMOS process. The physical layer signal processing for GSM, EDGE, GPRS, UMTS, HSPA, GMR1-3G, and LTE can be implemented in software on this single-chip device. Respective GSM, UMTS and LTE lab demonstrations have proven the competitiveness of this SDR approach.

We present initial results from the application of an automated facial expression recognition system to spontaneous facial expressions of pain. In this study, 26 participants were videotaped under three experimental conditions: baseline, posed pain, and real pain. In the real pain condition, subjects experienced cold pressor pain by submerging their arm in ice water. Our goal was to automatically determine which experimental condition was shown in a 60 second clip from a previously unseen subject. We chose a machine learning approach, previously used successfully to categorize basic emotional facial expressions in posed datasets as well as to detect individual facial actions of the Facial Action Coding System (FACS) (Littlewort et al, 2006; Bartlett et al., 2006). For this study, we trained 20 Action Unit (AU) classifiers on over 5000 images selected from a combination of posed and spontaneous facial expressions. The output of the system was a real valued number indicating the distance to the separating hyperplane for each classifier. Applying this system to the pain video data produced a 20 channel output stream, consisting of one real value for each learned AU, for each frame of the video. This data was passed to a second layer of classifiers to predict the difference between baseline and pained faces, and the difference between expressions of real pain and fake pain. Naíve human subjects tested on the same videos were at chance for differentiating faked from real pain, obtaining only 52% accuracy. The automated system was successfully able to differentiate faked from real pain. In an analysis of 26 subjects, the system obtained 72% correct for subject independent discrimination of real versus fake pain on a 2-alternative forced choice. Moreover, the most discriminative facial action in the automated system output was AU 4 (brow lower), which all was consistent with findings using human expert FACS codes.

A real-time compensation scheme for geometric and thermally-induced errors of a computerized numerical control (CNC) turning center is described. The compensation system predicts these errors using a combination of data taken from various sensors on the machine tool and previously established relationships (transfer functions). The system translates these errors into servo counts and injects them into the control loops of the machine tool controller in real-time. A single-board microcomputer interfaced to the machine tool controller and the sensors is the workhorse of the system. The system control software written in PLM, a high level programming language, is modular, flexible, and easily maintainable. To evaluate the capabilities of the compensation system, a series of cutting tests have been performed, and the results are presented. Along with the elimination of machine warmup periods of 8 to 12 hours, significant geometric accuracy improvements in straightness and squareness have been achieved. An additional accomplishment is a dimensional accuracy improvement of up to 20 times over that of uncompensated workpieces.