The recent availability of low-power neural accelerator hardware, combined with improvements in end-to-end neural facial recognition algorithms provides, enabling technology for on-device facial authentication. The present research work examines the effects of directional lighting on a State-of-Art(SoA) neural face recognizer. A synthetic re-lighting technique is used to augment data samples due to the lack of public data-sets with sufficient directional lighting variations. Top lighting and its variants (top-left, top-right) are found to have minimal effect on accuracy, while bottom-left or bottom-right directional lighting has the most pronounced effects. Following the fine-tuning of network weights, the face recognition model is shown to achieve close to the original Receiver Operating Characteristic curve (ROC)performance across all lighting conditions and demonstrates an ability to generalize beyond the lighting augmentations used in the fine-tuning data-set. This work shows that an SoA neural face recognition model can be tuned to compensate for directional lighting effects, removing the need for a pre-processing step before applying facial recognition.

I recently completed the Artificial Intelligence Product Manager Nanodegree Program on Udacity and I'd like to share a summary of everything I learned with you. This also includes bits from my experience as a technical product manager. This all a huge dump from my mind, written from the first stroke to last on my keyboard so kindly excuse any details I may miss or depths I didn't hit. It would be great to start with "why" and what motivated me to complete this program. In the past year, I've been working as a full-time product manager, sitting at the intersection of engineering and business and it's been fun. However, I'd recently been thinking deeply about the future of technology and what turns it could take.

The ROC curve is the gold standard for measuring the performance of a test/scoring statistic regarding its capacity to discriminate between two statistical populations in a wide variety of applications, ranging from anomaly detection in signal processing to information retrieval, through medical diagnosis. Most practical performance measures used in scoring/ranking applications such as the AUC, the local AUC, the p-norm push, the DCG and others, can be viewed as summaries of the ROC curve. In this paper, the fact that most of these empirical criteria can be expressed as two-sample linear rank statistics is highlighted and concentration inequalities for collections of such random variables, referred to as two-sample rank processes here, are proved, when indexed by VC classes of scoring functions. Based on these nonasymptotic bounds, the generalization capacity of empirical maximizers of a wide class of ranking performance criteria is next investigated from a theoretical perspective. It is also supported by empirical evidence through convincing numerical experiments.

We create an artificial system of agents (attention-based neural networks) which selectively exchange messages with each-other in order to study the emergence of memetic evolution and how memetic evolutionary pressures interact with genetic evolution of the network weights. We observe that the ability of agents to exert selection pressures on each-other is essential for memetic evolution to bootstrap itself into a state which has both high-fidelity replication of memes, as well as continuing production of new memes over time. However, in this system there is very little interaction between this memetic 'ecology' and underlying tasks driving individual fitness - the emergent meme layer appears to be neither helpful nor harmful to agents' ability to learn to solve tasks. Sourcecode for these experiments is available at https://github.com/GoodAI/memes

Because manufacturing processes evolve fast, and since production visual aspect can vary significantly on a daily basis, the ability to rapidly update machine vision based inspection systems is paramount. Unfortunately, supervised learning of convolutional neural networks requires a significant amount of annotated images for being able to learn effectively from new data. Acknowledging the abundance of continuously generated images coming from the production line and the cost of their annotation, we demonstrate it is possible to prioritize and accelerate the annotation process. In this work, we develop a methodology for learning actively, from rapidly mined, weakly (i.e. partially) annotated data, enabling a fast, direct feedback from the operators on the production line and tackling a big machine vision weakness: false positives. We also consider the problem of covariate shift, which arises inevitably due to changing conditions during data acquisition. In that regard, we show domain-adversarial training to be an efficient way to address this issue.

Electronic Health Record (EHR) data has been of tremendous utility in Artificial Intelligence (AI) for healthcare such as predicting future clinical events. These tasks, however, often come with many challenges when using classical machine learning models due to a myriad of factors including class imbalance and data heterogeneity (i.e., the complex intra-class variances). To address some of these research gaps, this paper leverages the exciting contrastive learning framework and proposes a novel contrastive regularized clinical classification model. The contrastive loss is found to substantially augment EHR-based prediction: it effectively characterizes the similar/dissimilar patterns (by its "push-and-pull" form), meanwhile mitigating the highly skewed class distribution by learning more balanced feature spaces (as also echoed by recent findings). In particular, when naively exporting the contrastive learning to the EHR data, one hurdle is in generating positive samples, since EHR data is not as amendable to data augmentation as image data. To this end, we have introduced two unique positive sampling strategies specifically tailored for EHR data: a feature-based positive sampling that exploits the feature space neighborhood structure to reinforce the feature learning; and an attribute-based positive sampling that incorporates pre-generated patient similarity metrics to define the sample proximity. Both sampling approaches are designed with an awareness of unique high intra-class variance in EHR data. Our overall framework yields highly competitive experimental results in predicting the mortality risk on real-world COVID-19 EHR data with a total of 5,712 patients admitted to a large, urban health system. Specifically, our method reaches a high AUROC prediction score of 0.959, which outperforms other baselines and alternatives: cross-entropy(0.873) and focal loss(0.931).

This paper introduces a novel dataset to help researchers evaluate their computer vision and audio models for accuracy across a diverse set of age, genders, apparent skin tones and ambient lighting conditions. Our dataset is composed of 3,011 subjects and contains over 45,000 videos, with an average of 15 videos per person. The videos were recorded in multiple U.S. states with a diverse set of adults in various age, gender and apparent skin tone groups. A key feature is that each subject agreed to participate for their likenesses to be used. Additionally, our age and gender annotations are provided by the subjects themselves. A group of trained annotators labeled the subjects' apparent skin tone using the Fitzpatrick skin type scale. Moreover, annotations for videos recorded in low ambient lighting are also provided. As an application to measure robustness of predictions across certain attributes, we provide a comprehensive study on the top five winners of the DeepFake Detection Challenge (DFDC). Experimental evaluation shows that the winning models are less performant on some specific groups of people, such as subjects with darker skin tones and thus may not generalize to all people. In addition, we also evaluate the state-of-the-art apparent age and gender classification methods. Our experiments provides a through analysis on these models in terms of fair treatment of people from various backgrounds.

Interpretability methods for image classification assess model trustworthiness by attempting to expose whether the model is systematically biased or attending to the same cues as a human would. Saliency methods for feature attribution dominate the interpretability literature, but these methods do not address semantic concepts such as the textures, colors, or genders of objects within an image. Our proposed Robust Concept Activation Vectors (RCAV) quantifies the effects of semantic concepts on individual model predictions and on model behavior as a whole. RCAV calculates a concept gradient and takes a gradient ascent step to assess model sensitivity to the given concept. By generalizing previous work on concept activation vectors to account for model non-linearity, and by introducing stricter hypothesis testing, we show that RCAV yields interpretations which are both more accurate at the image level and robust at the dataset level. RCAV, like saliency methods, supports the interpretation of individual predictions. To evaluate the practical use of interpretability methods as debugging tools, and the scientific use of interpretability methods for identifying inductive biases (e.g. texture over shape), we construct two datasets and accompanying metrics for realistic benchmarking of semantic interpretability methods. Our benchmarks expose the importance of counterfactual augmentation and negative controls for quantifying the practical usability of interpretability methods.

Machine Learning models have been deployed across almost every aspect of society, often in situations that affect the social welfare of many individuals. Although these models offer streamlined solutions to large problems, they may contain biases and treat groups or individuals unfairly. To our knowledge, this review is one of the first to focus specifically on gender bias in applications of machine learning. We first introduce several examples of machine learning gender bias in practice. We then detail the most widely used formalizations of fairness in order to address how to make machine learning models fairer. Specifically, we discuss the most influential bias mitigation algorithms as applied to domains in which models have a high propensity for gender discrimination. We group these algorithms into two overarching approaches -- removing bias from the data directly and removing bias from the model through training -- and we present representative examples of each. As society increasingly relies on artificial intelligence to help in decision-making, addressing gender biases present in these models is imperative. To provide readers with the tools to assess the fairness of machine learning models and mitigate the biases present in them, we discuss multiple open source packages for fairness in AI.

Detecting an Out-of-Domain (OOD) utterance is crucial for a robust dialog system. Most dialog systems are trained on a pool of annotated OOD data to achieve this goal. However, collecting the annotated OOD data for a given domain is an expensive process. To mitigate this issue, previous works have proposed generative adversarial networks (GAN) based models to generate OOD data for a given domain automatically. However, these proposed models do not work directly with the text. They work with the text's latent space instead, enforcing these models to include components responsible for encoding text into latent space and decoding it back, such as auto-encoder. These components increase the model complexity, making it difficult to train. We propose OodGAN, a sequential generative adversarial network (SeqGAN) based model for OOD data generation. Our proposed model works directly on the text and hence eliminates the need to include an auto-encoder. OOD data generated using OodGAN model outperforms state-of-the-art in OOD detection metrics for ROSTD (67% relative improvement in FPR 0.95) and OSQ datasets (28% relative improvement in FPR 0.95) (Zheng et al., 2020).