Nowadays, search ranking and recommendation systems rely on a lot of data to train machine learning models such as Learning-to-Rank (LTR) models to rank results for a given query, and implicit user feedbacks (e.g. click data) have become the dominant source of data collection due to its abundance and low cost, especially for major Internet companies. However, a drawback of this data collection approach is the data could be highly biased, and one of the most significant biases is the position bias, where users are biased towards clicking on higher ranked results. In this work, we will investigate the marginal importance of properly handling the position bias in an online test environment in Tripadvisor Hotels search. We propose an empirically effective method of handling the position bias that fully leverages the user action data. We take advantage of the fact that when user clicks a result, he has almost certainly observed all the results above, and the propensities of the results below the clicked result will be estimated by a simple but effective position bias model. The online A/B test results show that this method leads to an improved search ranking model.
The hypothesis that sub-network initializations (lottery) exist within the initializations of over-parameterized networks, which when trained in isolation produce highly generalizable models, has led to crucial insights into network initialization and has enabled computationally efficient inferencing. In order to realize the full potential of these pruning strategies, particularly when utilized in transfer learning scenarios, it is necessary to understand the behavior of winning tickets when they might overfit to the dataset characteristics. In supervised and semi-supervised learning, prediction calibration is a commonly adopted strategy to handle such inductive biases in models. In this paper, we study the impact of incorporating calibration strategies during model training on the quality of the resulting lottery tickets, using several evaluation metrics. More specifically, we incorporate a suite of calibration strategies to different combinations of architectures and datasets, and evaluate the fidelity of sub-networks retrained based on winning tickets. Furthermore, we report the generalization performance of tickets across distributional shifts, when the inductive biases are explicitly controlled using calibration mechanisms. Finally, we provide key insights and recommendations for obtaining reliable lottery tickets, which we demonstrate to achieve improved generalization.
Can effective international governance for artificial intelligence remain fragmented, or is there a need for a centralised international organisation for AI? We draw on the history of other international regimes to identify advantages and disadvantages in centralising AI governance. Some considerations, such as efficiency and political power, speak in favour of centralisation. Conversely, the risk of creating a slow and brittle institution speaks against it, as does the difficulty in securing participation while creating stringent rules. Other considerations depend on the specific design of a centralised institution. A well-designed body may be able to deter forum shopping and ensure policy coordination. However, forum shopping can be beneficial and a fragmented landscape of institutions can be self-organising. Centralisation entails trade-offs and the details matter. We conclude with two core recommendations. First, the outcome will depend on the exact design of a central institution. A well-designed centralised regime covering a set of coherent issues could be beneficial. But locking-in an inadequate structure may pose a fate worse than fragmentation. Second, for now fragmentation will likely persist. This should be closely monitored to see if it is self-organising or simply inadequate.
The self-organizing map (SOM) is an unsupervised artificial neural network that is widely used in, e.g., data mining and visualization. Supervised and semi-supervised learning methods have been proposed for the SOM. However, their teacher labels do not describe the relationship between the data and the location of nodes. This study proposes a landmark map (LAMA), which is an extension of the SOM that utilizes several landmarks, e.g., pairs of nodes and data points. LAMA is designed to obtain a user-intended nonlinear projection to achieve, e.g., the landmark-oriented data visualization. To reveal the learning properties of LAMA, the Zoo dataset from the UCI Machine Learning Repository and an artificial formant dataset were analyzed. The analysis results of the Zoo dataset indicated that LAMA could provide a new data view such as the landmark-centered data visualization. Furthermore, the artificial formant data analysis revealed that LAMA successfully provided the intended nonlinear projection associating articular movement with vertical and horizontal movement of a computer cursor. Potential applications of LAMA include data mining, recommendation systems, and human-computer interaction.
In this paper, we introduce a set of opinion annotations for the POM movie review dataset, composed of 1000 videos. The annotation campaign is motivated by the development of a hierarchical opinion prediction framework allowing one to predict the different components of the opinions (e.g. polarity and aspect) and to identify the corresponding textual spans. The resulting annotations have been gathered at two granularity levels: a coarse one (opinionated span) and a finer one (span of opinion components). We introduce specific categories in order to make the annotation of opinions easier for movie reviews. For example, some categories allow the discovery of user recommendation and preference in movie reviews. We provide a quantitative analysis of the annotations and report the inter-annotator agreement under the different levels of granularity. We provide thus the first set of ground-truth annotations which can be used for the task of fine-grained multimodal opinion prediction. We provide an analysis of the data gathered through an inter-annotator study and show that a linear structured predictor learns meaningful features even for the prediction of scarce labels. Both the annotations and the baseline system will be made publicly available.
Surrogate-based optimization and nature-inspired metaheuristics have become the state-of-the-art in solving real-world optimization problems. Still, it is difficult for beginners and even experts to get an overview that explains their advantages in comparison to the large number of available methods in the scope of continuous optimization. Available taxonomies lack the integration of surrogate-based approaches and thus their embedding in the larger context of this broad field. This article presents a taxonomy of the field, which further matches the idea of nature-inspired algorithms, as it is based on the human behavior in path finding. Intuitive analogies make it easy to conceive the most basic principles of the search algorithms, even for beginners and non-experts in this area of research. However, this scheme does not oversimplify the high complexity of the different algorithms, as the class identifier only defines a descriptive meta-level of the algorithm search strategies. The taxonomy was established by exploring and matching algorithm schemes, extracting similarities and differences, and creating a set of classification indicators to distinguish between five distinct classes. In practice, this taxonomy allows recommendations for the applicability of the corresponding algorithms and helps developers trying to create or improve their own algorithms.
Facial micro-expressions are very brief, spontaneous facial expressions that appear on the face of humans when they either deliberately or unconsciously conceal an emotion. Micro-expression has shorter duration than macro-expression, which makes it more challenging for human and machine. Over the past ten years, automatic micro-expressions recognition has attracted increasing attention from researchers in psychology, computer science, security, neuroscience and other related disciplines. The aim of this paper is to provide the insights of automatic micro-expressions and recommendations for future research. There has been a lot of datasets released over the last decade that facilitated the rapid growth in this field. However, comparison across different datasets is difficult due to the inconsistency in experiment protocol, features used and evaluation methods. To address these issues, we review the datasets, features and the performance metrics deployed in the literature. Relevant challenges such as the spatial temporal settings during data collection, emotional classes versus objective classes in data labelling, face regions in data analysis, standardisation of metrics and the requirements for real-world implementation are discussed. We conclude by proposing some promising future directions to advancing micro-expressions research.
Designing tax policies that are effective in curbing tax evasion and maximize state revenues requires a rigorous understanding of taxpayer behavior. This work explores the problem of determining the strategy a self-interested, risk-averse tax entity is expected to follow, as it "navigates" - in the context of a Markov Decision Process - a government-controlled tax environment that includes random audits, penalties and occasional tax amnesties. Although simplified versions of this problem have been previously explored, the mere assumption of risk-aversion (as opposed to risk-neutrality) raises the complexity of finding the optimal policy well beyond the reach of analytical techniques. Here, we obtain approximate solutions via a combination of Q-learning and recent advances in Deep Reinforcement Learning. By doing so, we i) determine the tax evasion behavior expected of the taxpayer entity, ii) calculate the degree of risk aversion of the "average" entity given empirical estimates of tax evasion, and iii) evaluate sample tax policies, in terms of expected revenues. Our model can be useful as a testbed for "in-vitro" testing of tax policies, while our results lead to various policy recommendations.
Building machines that can understand text like humans is an AI-complete problem. A great deal of research has already gone into this, with astounding results, allowing everyday people to discuss with their telephones, or have their reading materials analysed and classified by computers. A prerequisite for processing text semantics, common to the above examples, is having some computational representation of text as an abstract object. Operations on this representation practically correspond to making semantic inferences, and by extension simulating understanding text. The complexity and granularity of semantic processing that can be realised is constrained by the mathematical and computational robustness, expressiveness, and rigour of the tools used. This dissertation contributes a series of such tools, diverse in their mathematical formulation, but common in their application to model semantic inferences when machines process text. These tools are principally expressed in nine distinct models that capture aspects of semantic dependence in highly interpretable and non-complex ways. This dissertation further reflects on present and future problems with the current research paradigm in this area, and makes recommendations on how to overcome them. The amalgamation of the body of work presented in this dissertation advances the complexity and granularity of semantic inferences that can be made automatically by machines.
Collaborative filtering (CF) and content-based filtering (CBF) have widely been used in information filtering applications. Both approaches have their strengths and weaknesses which is why researchers have developed hybrid systems. This paper proposes a novel approach to unify CF and CBF in a probabilistic framework, named collaborative ensemble learning. It uses probabilistic SVMs to model each user's profile (as CBF does).At the prediction phase, it combines a society OF users profiles, represented by their respective SVM models, to predict an active users preferences(the CF idea).The combination scheme is embedded in a probabilistic framework and retains an intuitive explanation.Moreover, collaborative ensemble learning does not require a global training stage and thus can incrementally incorporate new data.We report results based on two data sets. For the Reuters-21578 text data set, we simulate user ratings under the assumption that each user is interested in only one category. In the second experiment, we use users' opinions on a set of 642 art images that were collected through a web-based survey. For both data sets, collaborative ensemble achieved excellent performance in terms of recommendation accuracy.