Artificial Intelligence (AI) or Machine Learning (ML) systems have been widely adopted as value propositions by companies in all industries in order to create or extend the services and products they offer. However, developing AI/ML systems has presented several engineering problems that are different from those that arise in, non-AI/ML software development. This study aims to investigate how software engineering (SE) has been applied in the development of AI/ML systems and identify challenges and practices that are applicable and determine whether they meet the needs of professionals. Also, we assessed whether these SE practices apply to different contexts, and in which areas they may be applicable. We conducted a systematic review of literature from 1990 to 2019 to (i) understand and summarize the current state of the art in this field and (ii) analyze its limitations and open challenges that will drive future research. Our results show these systems are developed on a lab context or a large company and followed a research-driven development process. The main challenges faced by professionals are in areas of testing, AI software quality, and data management. The contribution types of most of the proposed SE practices are guidelines, lessons learned, and tools.

Synthesizing realistic data samples is of great value for both academic and industrial communities. Deep generative models have become an emerging topic in various research areas like computer vision and signal processing. Affective computing, a topic of a broad interest in computer vision society, has been no exception and has benefited from generative models. In fact, affective computing observed a rapid derivation of generative models during the last two decades. Applications of such models include but are not limited to emotion recognition and classification, unimodal emotion synthesis, and cross-modal emotion synthesis. As a result, we conducted a review of recent advances in human emotion synthesis by studying available databases, advantages, and disadvantages of the generative models along with the related training strategies considering two principal human communication modalities, namely audio and video. In this context, facial expression synthesis, speech emotion synthesis, and the audio-visual (cross-modal) emotion synthesis is reviewed extensively under different application scenarios. Gradually, we discuss open research problems to push the boundaries of this research area for future works.

This work represents a natural coalescence of two important lines of work: learning mixtures of Gaussians and algorithmic robust statistics. In particular we give the first provably robust algorithm for learning mixtures of any constant number of Gaussians. We require only mild assumptions on the mixing weights (bounded fractionality) and that the total variation distance between components is bounded away from zero. At the heart of our algorithm is a new method for proving dimension-independent polynomial identifiability through applying a carefully chosen sequence of differential operations to certain generating functions that not only encode the parameters we would like to learn but also the system of polynomial equations we would like to solve. We show how the symbolic identities we derive can be directly used to analyze a natural sum-of-squares relaxation.

Due to accessible big data collections from consumers, products, and stores, advanced sales forecasting capabilities have drawn great attention from many companies especially in the retail business because of its importance in decision making. Improvement of the forecasting accuracy, even by a small percentage, may have a substantial impact on companies' production and financial planning, marketing strategies, inventory controls, supply chain management, and eventually stock prices. Specifically, our research goal is to forecast the sales of each product in each store in the near future. Motivated by tensor factorization methodologies for personalized context-aware recommender systems, we propose a novel approach called the Advanced Temporal Latent-factor Approach to Sales forecasting (ATLAS), which achieves accurate and individualized prediction for sales by building a single tensor-factorization model across multiple stores and products. Our contribution is a combination of: tensor framework (to leverage information across stores and products), a new regularization function (to incorporate demand dynamics), and extrapolation of tensor into future time periods using state-of-the-art statistical (seasonal auto-regressive integrated moving-average models) and machine-learning (recurrent neural networks) models. The advantages of ATLAS are demonstrated on eight product category datasets collected by the Information Resource, Inc., where a total of 165 million weekly sales transactions from more than 1,500 grocery stores over 15,560 products are analyzed.

One Class Slab Support Vector Machines (OCSSVM) have turned out to be better in terms of accuracy in certain classes of classification problems than the traditional SVMs and One Class SVMs or even other One class classifiers. This paper proposes fast training method for One Class Slab SVMs using an updated Sequential Minimal Optimization (SMO) which divides the multi variable optimization problem to smaller sub problems of size two that can then be solved analytically. The results indicate that this training method scales better to large sets of training data than other Quadratic Programming (QP) solvers.

The ever-growing advances of deep learning in many areas including vision, recommendation systems, natural language processing, etc., have led to the adoption of Deep Neural Networks (DNNs) in production systems. The availability of large datasets and high computational power are the main contributors to these advances. The datasets are usually crowdsourced and may contain sensitive information. This poses serious privacy concerns as this data can be misused or leaked through various vulnerabilities. Even if the cloud provider and the communication link is trusted, there are still threats of inference attacks where an attacker could speculate properties of the data used for training, or find the underlying model architecture and parameters. In this survey, we review the privacy concerns brought by deep learning, and the mitigating techniques introduced to tackle these issues. We also show that there is a gap in the literature regarding test-time inference privacy, and propose possible future research directions.

A major cause of uncertainties in climate projections is our imprecise knowledge of how much warming should occur as a result of a given increase in the amount of carbon dioxide in the atmosphere. Paleoclimate records have the potential to help us sharpen that understanding because they record such a wide variety of environmental conditions. Tierney et al. review the recent advances in data collection, statistics, and modeling that might help us better understand how rising levels of atmospheric carbon dioxide will affect future climate. Science , this issue p. [eaay3701][1] ### BACKGROUND Anthropogenic emissions are rapidly altering Earth’s climate, pushing it toward a warmer state for which there is no historical precedent. Although no perfect analog exists for such a disruption, Earth’s history includes past climate states—“paleoclimates”—that hold lessons for the future of our warming world. These periods in Earth’s past span a tremendous range of temperatures, precipitation patterns, cryospheric extent, and biospheric adaptations and are increasingly relevant for improving our understanding of how key elements of the climate system are affected by greenhouse gas levels. The rise of new geochemical and statistical methods, as well as improvements in paleoclimate modeling, allow for formal evaluation of climate models based on paleoclimate data. In particular, given that some of the newest generation of climate models have a high sensitivity to a doubling of atmospheric CO2, there is a renewed role for paleoclimates in constraining equilibrium climate sensitivity (ECS) and its dependence on climate background state. ### ADVANCES In the past decade, an increasing number of studies have used paleoclimate temperature and CO2 estimates to infer ECS in the deep past, in both warm and cold climate states. Recent studies support the paradigm that ECS is strongly state-dependent, rising with increased CO2 concentrations. Simulations of past warm climates such as the Eocene further highlight the role that cloud feedbacks play in contributing to high ECS under increased CO2 levels. Paleoclimates have provided critical constraints on the assessment of future ice sheet stability and concomitant sea level rise, including the viability of threshold processes like marine ice cliff instability. Beyond global-scale changes, analyses of past changes in the water cycle have advanced our understanding of dynamical drivers of hydroclimate, which is highly relevant for regional climate projections and societal impacts. New and expanding techniques, such as analyses of single shells of foraminifera, are yielding subseasonal climate information that can be used to study how intra- and interannual modes of variability are affected by external climate forcing. Studies of extraordinary, transient departures in paleoclimate from the background state such as the Paleocene-Eocene Thermal Maximum provide critical context for the current anthropogenic aberration, its impact on the Earth system, and the time scale of recovery. A number of advances have eroded the “language barrier” between climate model and proxy data, facilitating more direct use of paleoclimate information to constrain model performance. It is increasingly common to incorporate geochemical tracers, such as water isotopes, directly into model simulations, and this practice has vastly improved model-proxy comparisons. The development of new statistical approaches rooted in Bayesian inference has led to a more thorough quantification of paleoclimate data uncertainties. In addition, techniques like data assimilation allow for a formal combination of proxy and model data into hybrid products. Such syntheses provide a full-field view of past climates and can put constraints on climate variables that we have no direct proxies for, such as cloud cover or wind speed. ### OUTLOOK A common concern with using paleoclimate information as model targets is that non-CO2 forcings, such as aerosols and trace greenhouse gases, are not well known, especially in the distant past. Although evidence thus far suggests that such forcings are secondary to CO2, future improvements in both geochemical proxies and modeling are on track to tackle this issue. New and rapidly evolving geochemical techniques have the potential to provide improved constraints on the terrestrial biosphere, aerosols, and trace gases; likewise, biogeochemical cycles can now be incorporated into paleoclimate model simulations. Beyond constraining forcings, it is critical that proxy information is transformed into quantitative estimates that account for uncertainties in the proxy system. Statistical tools have already been developed to achieve this, which should make it easier to create robust targets for model evaluation. With this increase in quantification of paleoclimate information, we suggest that modeling centers include simulation of past climates in their evaluation and statement of their model performance. This practice is likely to narrow uncertainties surrounding climate sensitivity, ice sheets, and the water cycle and thus improve future climate projections. ![Figure][2] Past climates provide context for future climate scenarios. Both past (top) and future (bottom) climates are colored by their estimated change in global mean annual surface temperature relative to preindustrial conditions, ranging from blue (colder) to red (warmer). “Sustainability,” “Middle road,” and “High emissions” represent the estimated global temperature anomalies at year 2300 from the Shared Socioeconomic Pathways (SSPs) SSP1-2.6, SSP2-4.5, and SSP5-8.5, respectively. In both the past and future cases, warmer climates are associated with increases in CO2 (indicated by the arrow). Ma, millions of years ago. As the world warms, there is a profound need to improve projections of climate change. Although the latest Earth system models offer an unprecedented number of features, fundamental uncertainties continue to cloud our view of the future. Past climates provide the only opportunity to observe how the Earth system responds to high carbon dioxide, underlining a fundamental role for paleoclimatology in constraining future climate change. Here, we review the relevancy of paleoclimate information for climate prediction and discuss the prospects for emerging methodologies to further insights gained from past climates. Advances in proxy methods and interpretations pave the way for the use of past climates for model evaluation—a practice that we argue should be widely adopted. [1]: /lookup/doi/10.1126/science.aay3701 [2]: pending:yes

AGI or'Strong AI' is the concept of a machine with general intelligence that mimics human intelligence and/or behaviors, with the ability to learn and apply its intelligence to solve any problem. It possesses the ability to think and act in a way not any different from the way we humans think or act. We humans can drive a car, classify a pet by looking at its image, and grade an essay all with just one brain of ours. While an ANI system cannot do so, an AGI can! True AGI has not been achieved yet by AI researchers and scientists, owing to intricacies and complexities that it proposes. Who knows, maybe someday technological advances like Quantum Computers can enable achieve True AGI!

We address the problem of building theoretical models that help elucidate the function of the visual brain at computational/algorithmic and structural/mechanistic levels. We seek to understand how the receptive fields and topographic maps found in visual cortical areas relate to underlying computational desiderata. We view the development of sensory systems from the popular perspective of probability density estimation; this is motivated by the notion that an effective internal representational scheme is likely to reflect the statistical structure of the environment in which an organism lives. We apply biologically based constraints on elements of the model. The thesis begins by surveying the relevant literature from the fields of neurobiology, theoretical neuroscience, and machine learning. After this review we present our main theoretical and algorithmic developments: we propose a class of probabilistic models, which we refer to as "energy-based models", and show equivalences between this framework and various other types of probabilistic model such as Markov random fields and factor graphs; we also develop and discuss approximate algorithms for performing maximum likelihood learning and inference in our energy based models. The rest of the thesis is then concerned with exploring specific instantiations of such models. By performing constrained optimisation of model parameters to maximise the likelihood of appropriate, naturalistic datasets we are able to qualitatively reproduce many of the receptive field and map properties found in vivo, whilst simultaneously learning about statistical regularities in the data.

When an agent interacts with a complex environment, it receives a stream of percepts in which it may detect entities, such as objects or people. To build up a coherent, low-variance estimate of the underlying state, it is necessary to fuse information from multiple detections over time. To do this fusion, the agent must decide which detections to associate with one another. We address this data-association problem in the setting of an online filter, in which each observation is processed by aggregating into an existing object hypothesis. Classic methods with strong probabilistic foundations exist, but they are computationally expensive and require models that can be difficult to acquire. In this work, we use the deep-learning tools of sparse attention and representation learning to learn a machine that processes a stream of detections and outputs a set of hypotheses about objects in the world. We evaluate this approach on simple clustering problems, problems with dynamics, and a complex image-based domain. We find that it generalizes well from short to long observation sequences and from a few to many hypotheses, outperforming other learning approaches and classical non-learning methods.