Offline reinforcement learning (RL) tasks require the agent to learn from a pre-collected dataset with no further interactions with the environment. Despite the potential to surpass the behavioral policies, RL-based methods are generally impractical due to the training instability and bootstrapping the extrapolation errors, which always require careful hyperparameter tuning via online evaluation. In contrast, offline imitation learning (IL) has no such issues since it learns the policy directly without estimating the value function by bootstrapping. However, IL is usually limited in the capability of the behavioral policy and tends to learn a mediocre behavior from the dataset collected by the mixture of policies. In this paper, we aim to take advantage of IL but mitigate such a drawback. Observing that behavior cloning is able to imitate neighboring policies with less data, we propose \textit{Curriculum Offline Imitation Learning (COIL)}, which utilizes an experience picking strategy for imitating from adaptive neighboring policies with a higher return, and improves the current policy along curriculum stages. On continuous control benchmarks, we compare COIL against both imitation-based and RL-based methods, showing that it not only avoids just learning a mediocre behavior on mixed datasets but is also even competitive with state-of-the-art offline RL methods.

WENDELL WALLACH: It gives me great pleasure to welcome my longtime colleague Shannon Vallor to this Artificial Intelligence & Equality podcast. Shannon and I have both expressed concerns that ethics and ethical philosophy is inadequate for addressing the issues posed by artificial intelligence (AI) and other emerging technologies, so I have been looking forward to our having a conversation about why that is the case and ideas for reenvisioning ethics and empowering it for the information age. Before we get to that conversation, let me introduce Shannon to our listeners, provide a very cursory overview of how ethical theories are understood within academic circles, and provide Shannon with the opportunity to introduce you to the research and insights for which she is best known. Again, before turning to Shannon, let me make sure that listeners have at least a cursory understanding of the field of ethics. Ethical theories are often said to all fall into two big tents, and one of those tents--the determination of what is right, good, or just--derives from following the rules or doing your duty. Often these rules are captured in high-level principles, that the rules can be the Ten Commandments or the four principles of biomedical ethics. In India they might be Yama and Niyama. Each culture has its own set of rules. Even Asimov's "Three Laws of Robotics" do count as rules meant to direct the behavior of robots. All of these theories are said to be deontological, a term going back to the Greeks, referring to duties, and it is basically saying that rules and duties define ethics--but of course there are outstanding questions about whose rules, what to do when rules conflict, and how you deal with situations when people prioritize the rules very differently. At the end of the 18th and beginning of the 19th centuries, Jeremy Bentham, a British philosopher, came up with a totally different approach to ethics, which is sometimes called utilitarianism or consequentialism.

Machine learning becomes engaging when we face various challenges and thus finding suitable datasets relevant to the use case is essential. Flexibility refers to the number of tasks that it supports. For example, Microsoft's COCO( Common Objects in Context) is used for object classification, detection, and segmentation. Add a bunch of captions for the same, and we can use it as a dataset for an image caption generator as well. Well, when we are just starting, we shall be working with some of the small and standard machine learning datasets like the CIFAR-10, MNIS, Iris, etc.

Unobserved confounding is the main obstacle to causal effect estimation from observational data. Instrumental variables (IVs) are widely used for causal effect estimation when there exist latent confounders. With the standard IV method, when a given IV is valid, unbiased estimation can be obtained, but the validity requirement of standard IV is strict and untestable. Conditional IV has been proposed to relax the requirement of standard IV by conditioning on a set of observed variables (known as a conditioning set for a conditional IV). However, the criterion for finding a conditioning set for a conditional IV needs complete causal structure knowledge or a directed acyclic graph (DAG) representing the causal relationships of both observed and unobserved variables. This makes it impossible to discover a conditioning set directly from data. In this paper, by leveraging maximal ancestral graphs (MAGs) in causal inference with latent variables, we propose a new type of IV, ancestral IV in MAG, and develop the theory to support data-driven discovery of the conditioning set for a given ancestral IV in MAG. Based on the theory, we develop an algorithm for unbiased causal effect estimation with an ancestral IV in MAG and observational data. Extensive experiments on synthetic and real-world datasets have demonstrated the performance of the algorithm in comparison with existing IV methods.

Real-world machine learning deployments are characterized by mismatches between the source (training) and target (test) distributions that may cause performance drops. In this work, we investigate methods for predicting the target domain accuracy using only labeled source data and unlabeled target data. We propose Average Thresholded Confidence (ATC), a practical method that learns a threshold on the model's confidence, predicting accuracy as the fraction of unlabeled examples for which model confidence exceeds that threshold. ATC outperforms previous methods across several model architectures, types of distribution shifts (e.g., due to synthetic corruptions, dataset reproduction, or novel subpopulations), and datasets (W In our experiments, ATC estimates target performance 2-4ˆ more accurately than prior methods. We also explore the theoretical foundations of the problem, proving that, in general, identifying the accuracy is just as hard as identifying the optimal predictor and thus, the efficacy of any method rests upon (perhaps unstated) assumptions on the nature of the shift. Finally, analyzing our method on some toy distributions, we provide insights concerning when it works. Machine learning models deployed in the real world typically encounter examples from previously unseen distributions. While the IID assumption enables us to evaluate models using held-out data from the source distribution (from which training data is sampled), this estimate is no longer valid in presence of a distribution shift. Moreover, under such shifts, model accuracy tends to degrade (Szegedy et al., 2014; Recht et al., 2019; Koh et al., 2021). Commonly, the only data available to the practitioner are a labeled training set (source) and unlabeled deployment-time data which makes the problem more difficult. In this setting, detecting shifts in the distribution of covariates is known to be possible (but difficult) in theory (Ramdas et al., 2015), and in practice (Rabanser et al., 2018). However, producing an optimal predictor using only labeled source and unlabeled target data is well-known to be impossible absent further assumptions (Ben-David et al., 2010; Lipton et al., 2018). Two vital questions that remain are: (i) the precise conditions under which we can estimate a classifier's target-domain accuracy; and (ii) which methods are most practically useful. To begin, the straightforward way to assess the performance of a model under distribution shift would be to collect labeled (target domain) examples and then to evaluate the model on that data. However, collecting fresh labeled data from the target distribution is prohibitively expensive and time-consuming, especially if the target distribution is non-stationary.

The detection of hate speech online has become an important task, as offensive language such as hurtful, obscene and insulting content can harm marginalized people or groups. This paper presents TU Berlin team experiments and results on the task 1A and 1B of the shared task on hate speech and offensive content identification in Indo-European languages 2021. The success of different Natural Language Processing models is evaluated for the respective subtasks throughout the competition. We tested different models based on recurrent neural networks in word and character levels and transfer learning approaches based on Bert on the provided dataset by the competition. Among the tested models that have been used for the experiments, the transfer learning-based models achieved the best results in both subtasks.

Recent works explore learning graph representations in a self-supervised manner. In graph contrastive learning, benchmark methods apply various graph augmentation approaches. However, most of the augmentation methods are non-learnable, which causes the issue of generating unbeneficial augmented graphs. Such augmentation may degenerate the representation ability of graph contrastive learning methods. Therefore, we motivate our method to generate augmented graph by a learnable graph augmenter, called MEta Graph Augmentation (MEGA). We then clarify that a "good" graph augmentation must have uniformity at the instance-level and informativeness at the feature-level. To this end, we propose a novel approach to learning a graph augmenter that can generate an augmentation with uniformity and informativeness. The objective of the graph augmenter is to promote our feature extraction network to learn a more discriminative feature representation, which motivates us to propose a meta-learning paradigm. Empirically, the experiments across multiple benchmark datasets demonstrate that MEGA outperforms the state-of-the-art methods in graph self-supervised learning tasks. Further experimental studies prove the effectiveness of different terms of MEGA.

This paper reports the results and post-challenge analyses of ChaLearn's AutoDL challenge series, which helped sorting out a profusion of AutoML solutions for Deep Learning (DL) that had been introduced in a variety of settings, but lacked fair comparisons. All input data modalities (time series, images, videos, text, tabular) were formatted as tensors and all tasks were multi-label classification problems. Code submissions were executed on hidden tasks, with limited time and computational resources, pushing solutions that get results quickly. In this setting, DL methods dominated, though popular Neural Architecture Search (NAS) was impractical. Solutions relied on fine-tuned pre-trained networks, with architectures matching data modality. Post-challenge tests did not reveal improvements beyond the imposed time limit. While no component is particularly original or novel, a high level modular organization emerged featuring a "meta-learner", "data ingestor", "model selector", "model/learner", and "evaluator". This modularity enabled ablation studies, which revealed the importance of (off-platform) meta-learning, ensembling, and efficient data management. Experiments on heterogeneous module combinations further confirm the (local) optimality of the winning solutions. Our challenge legacy includes an ever-lasting benchmark (http://autodl.chalearn.org), the open-sourced code of the winners, and a free "AutoDL self-service".

Mike is a 40-something crop farmer from southern Queensland. With a chestnut tan, crushing handshake and a strong outback accent, he's the third generation of his family to grow sorghum, a cereal mostly used for animal fodder. But, like most farmers, Mike faces more challenges than his forbears. Climate change has eroded Australian farms' profitability by an average of 23% over the past 20 years. It's a constant challenge to improve productivity by producing more with less.

Community detection is one of the most important methodological fields of network science, and one which has attracted a significant amount of attention over the past decades. This area deals with the automated division of a network into fundamental building blocks, with the objective of providing a summary of its large-scale structure. Despite its importance and widespread adoption, there is a noticeable gap between what is considered the state-of-the-art and the methods that are actually used in practice in a variety of fields. Here we attempt to address this discrepancy by dividing existing methods according to whether they have a "descriptive" or an "inferential" goal. While descriptive methods find patterns in networks based on intuitive notions of community structure, inferential methods articulate a precise generative model, and attempt to fit it to data. In this way, they are able to provide insights into the mechanisms of network formation, and separate structure from randomness in a manner supported by statistical evidence. We review how employing descriptive methods with inferential aims is riddled with pitfalls and misleading answers, and thus should be in general avoided. We argue that inferential methods are more typically aligned with clearer scientific questions, yield more robust results, and should be in many cases preferred. We attempt to dispel some myths and half-truths often believed when community detection is employed in practice, in an effort to improve both the use of such methods as well as the interpretation of their results.