Scientific papers

Martin Wistuba; Josif Grabocka;

Abstract:

Hyperparameter optimization (HPO) is a central pillar in the automation of machine learning solutions and is mainly performed via Bayesian optimization, where a parametric surrogate is learned to approximate the black box response function (e.g. validation error). Unfortunately, evaluating the response function is computationally intensive. As a remedy, earlier work emphasizes the need for transfer learning surrogates which learn to optimize hyperparameters for an algorithm from other tasks. In contrast to previous work, we propose to rethink HPO as a few-shot learning problem in which we train a shared deep surrogate model to quickly adapt (with few response evaluations) to the response function of a new task. We propose the use of a deep kernel network for a Gaussian process surrogate that is meta-learned in an end-to-end fashion in order to jointly approximate the response functions of a collection of training data sets. As a result, the novel few-shot optimization of our deep kernel surrogate leads to new state-of-the-art results at HPO compared to several recent methods on diverse metadata sets.

Moreo, Alejandro; Sebastiani, Fabrizio

Abstract:

Learning to quantify (a.k.a. quantification) is a task concerned with training unbiased estimators of class prevalence via supervised learning. This task originated with the observation that “Classify and Count” (CC), the trivial method of obtaining class prevalence estimates, is often a biased estimator, and thus delivers suboptimal quantification accuracy. Fol- lowing this observation, several methods for learning to quantify have been proposed and have been shown to outperform CC. In this work we contend that previous works have failed to use properly optimised versions of CC. We thus reassess the real merits of CC and its variants, and argue that, while still inferior to some cutting-edge methods, they deliver near-state-of-the- art accuracy once (a) hyperparameter optimisation is performed, and (b) this optimisation is performed by using a truly quantification-oriented evaluation protocol. Experiments on three publicly available binary sentiment classification datasets support these conclusions.

Moreo, Alejandro; Pedrotti, Andrea; Sebastiani, Fabrizio

Abstract:

Funnelling (Fun) is a method for cross-lingual text classification (CLC) based on a two-tier ensemble for heterogeneous transfer learning. In Fun, 1st-tier classifiers, each working on a different, language-dependent feature space, return a vector of calibrated posterior probabilities (with one dimension for each class) for each document, and the final classification decision is taken by a meta- classifier that uses this vector as its input. The metaclassifier can thus exploit class-class correlations, and this (among other things) gives Fun an edge over CLC systems where these correlations cannot be leveraged.

We here describe Generalized Funnelling (gFun), a learning ensemble where the metaclassifier receives as input the above vector of calibrated posterior probabilities, concatenated with document embeddings (aligned across languages) that embody other types of correlations, such as word-class correlations (as encoded by Word-Class Embeddings) and word-word correlations (as encoded by Multilingual Unsupervised or Supervised Embeddings). We show that gFun improves on Fun by describing experiments on two large, standard multilingual datasets for multi-label text classification.

Mara Graziani; Thomas Lompech; Henning Müller; Vincent Andrearczyk;

Abstract:

Visualization methods for Convolutional Neural Net-works (CNNs) are spreading within the medical com-munity to obtain explainable AI (XAI). The sole quali-tative assessment of the explanations is subject to a riskof confirmation bias. This paper proposes a methodol-ogy for the quantitative evaluation of common visual-ization approaches for histopathology images, i.e. ClassActivation Mapping and Local-Interpretable Model-Agnostic Explanations. In our evaluation, we proposeto assess four main points, namely the alignment withclinical factors, the agreement between XAI methods,the consistency and repeatability of the explanations. Todo so, we compare the intersection over union of multi-ple visualizations of the CNN attention with the seman-tic annotation of functionally different nuclei types. Theexperimental results do not show stronger attributions tothe multiple nuclei types than those of a randomly ini-tialized CNN. The visualizations hardly agree on salientareas and LIME outputs have particularly unstable re-peatability and consistency. The qualitative evaluationalone is thus not sufficient to establish the appropriate-ness and reliability of the visualization tools. The codeis available on GitHub atbit.ly/2K48HKz.

Moreo, Alejandro; Esuli, Andrea; Sebastiani, Fabrizio;

Abstract:

Pre-trained word embeddings encode general word semantics and lexical regularities of natural language, and have proven useful across many NLP tasks, including word sense disambiguation, machine translation, and sentiment analysis, to name a few. In supervised tasks such as multiclass text classification (the focus of this article) it seems appealing to enhance word representations with ad-hoc embeddings that encode task-specific information. We propose (supervised) word-class embeddings (WCEs), and show that, when concatenated to (unsupervised) pre-trained word embeddings, they substantially facilitate the training of deep-learning models in multiclass classification by topic. We show empirical evidence that WCEs yield a consistent improvement in multiclass classification accuracy, using six popular neural architectures and six widely used and publicly available datasets for multi- class text classification. One further advantage of this method is that it is conceptually simple and straightforward to implement. Our code that implements WCEs is publicly available at https://github.com/AlexMoreo/ word-class-embeddings.

Esuli, Andrea; Molinari, Alessio; Sebastiani, Fabrizio;

Abstract:

We critically re-examine the Saerens-Latinne-Decaestecker (SLD) algorithm, a well-known method for estimating class prior probabilities (“priors”) and adjusting posterior probabilities (“posteriors”) in scenarios characterized by distribution shift, i.e., difference in the distribution of the priors between the training and the unlabelled documents. Given a machine learned classifier and a set of unlabelled documents for which the classifier has returned posterior probabilities and estimates of the prior probabilities, SLD updates them both in an iterative, mutually recursive way, with the goal of making both more accurate; this is of key importance in downstream tasks such as single-label multiclass classification and cost-sensitive text classification. Since its publication, SLD has become the standard algorithm for improving the quality of the posteriors in the presence of distribution shift, and SLD is still considered a top contender when we need to estimate the priors (a task that has become known as “quantification”). However, its real effectiveness in improving the quality of the posteriors has been questioned. We here present the results of systematic experiments conducted on a large, publicly available dataset, across multiple amounts of distribution shift and multiple learners. Our experiments show that SLD improves the quality of the posterior probabilities and of the estimates of the prior probabilities, but only when the number of classes in the classification scheme is very small and the classifier is calibrated. As the number of classes grows, or as we use non-calibrated classifiers, SLD converges more slowly (and often does not converge at all), performance degrades rapidly, and the impact of SLD on the quality of the prior estimates and of the posteriors becomes negative rather than positive.

Belouadah, Eden; Popescu, Adrian; Kanellos, Ioannis

Abstract:

The ability of artificial agents to increment their capabilities when confronted with new data is an open challenge in artificial intelligence. The main challenge faced in such cases is catastrophic forgetting, i.e., the tendency of neural networks to underfit past data when new ones are ingested. A first group of approaches tackles forgetting by increasing deep model capacity to accommodate new knowledge. A second type of approaches fix the deep model size and introduce a mechanism whose objective is to ensure a good compromise between stability and plasticity of the model. While the first type of algorithms were compared thoroughly, this is not the case for methods which exploit a fixed size model. Here, we focus on the latter, place them in a common conceptual and experimental framework and propose the following contributions: (1) define six desirable properties of incremental learning algorithms and analyze them according to these properties, (2) introduce a unified formalization of the class-incremental learning problem, (3) propose a common evaluation framework which is more thorough than existing ones in terms of number of datasets, size of datasets, size of bounded memory and number of incremental states, (4) investigate the usefulness of herding for past exemplars selection, (5) provide experimental evidence that it is possible to obtain competitive performance without the use of knowledge distillation to tackle catastrophic forgetting and (6) facilitate reproducibility by integrating all tested methods in a common open-source repository. The main experimental finding is that none of the existing algorithms achieves the best results in all evaluated settings. Important differences arise notably if a bounded memory of past classes is allowed or not.

Jialin Liu; Sam Snodgrass; Ahmed Khalifa; Sebastian Risi; Georgios N. Yannakakis; Julian Togelius

Abstract:

Procedural content generation in video games has a long history. Existing procedural content generation methods, such as search-based, solver-based, rule-based and grammar-based methods have been applied to various content types such as levels, maps, character models, and textures. A research field centered on content generation in games has existed for more than a decade. More recently, deep learning has powered a remarkable range of inventions in content production, which are applicable to games. While some cutting-edge deep learning methods are applied on their own, others are applied in combination with more traditional methods, or in an interactive setting. This article surveys the various deep learning methods that have been applied to generate game content directly or indirectly, discusses deep learning methods that could be used for content generation purposes but are rarely used today, and envisages some limitations and potential future directions of deep learning for procedural content generation.