Neural Information Processing Systems http://nips.cc/

We categorize meta-learning evaluation into two settings: in-distribution [ID], in which the train and test tasks are sampled iid from the same underlying task distribution, and out-of-distribution [OOD], in which they are not. While most metalearning theory and some FSL applications follow the ID setting, we identify that most existing few-shot classification benchmarks instead reflect OOD evaluation, as they use disjoint sets of train (base) and test (novel) classes for task generation. This discrepancy is problematic because--as we show on numerous benchmarks-- meta-learning methods that perform better on existing OOD datasets may perform significantly worse in the ID setting. In addition, in the OOD setting, even though current FSL benchmarks seem befitting, our study highlights concerns in 1) reliably performing model selection for a given meta-learning method, and 2) consistently comparing the performance of different methods. To address these concerns, we provide suggestions on how to construct FSL benchmarks to allow for ID evaluation as well as more reliable OOD evaluation. Our work aims to inform the meta-learning community about the importance and distinction of ID vs. OOD evaluation, as well as the subtleties of OOD evaluation with current benchmarks.

The dataset and the baseline code will be made publicly available in a dedicated GitHub repository upon acceptance. License TempEL is distributed under Creative Commons Attribution-ShareAlike 4.0 International license (CCBY-SA 4.0).1 Maintenance The maintenance and extension to further temporal snapshots of TempEL will be carried out by the authors of the paper. Additionally, we will make the code public to create potential new variations and extensions of TempEL using a number of hyperparameters (see Sections A.4 and A.5 for further details). A.2 Datasheet for TempEL In this section we provide a more detailed documentation of the dataset with the intended uses. We base ourselves on the datasheet proposed by [1]. A.2.1 Motivation For what purpose was the dataset created? The TempEL dataset was created to evaluate how the temporal change of anchor mentions and that of target Knowledge Base (KB; i.e., modification or creation of new entities) affects the entity linking (EL) task. This contrasts with the currently existing datasets [9, 7, 8, 6], which are associated with a single version of the target KB such as the Wikipedia 2010 for the widely adopted CoNLL-AIDA[2] dataset. We expect that TempEL will encourage research in devising new models and architectures that are robust to temporal changes both in mentions as well as in the target KBs. Who created the dataset and on behalf of which entity?

We investigate the parameterized complexity of Bayesian Network Structure Learn-1 ing (BNSL), a classical problem that has received significant attention in empirical2 but also purely theoretical studies. We follow up on previous works that have3 analyzed the complexity of BNSL w.r.t. the so-called superstructure of the input.4 While known results imply that BNSL is unlikely to be fixed-parameter tractable5 even when parameterized by the size of a vertex cover in the superstructure, here we6 show that a different kind of parameterization--notably by the size of a feedback7 edge set--yields fixed-parameter tractability. We proceed by showing that this8 result can be strengthened to a localized version of the feedback edge set, and9 provide corresponding lower bounds that complement previous results to provide a10 complexity classification of BNSL w.r.t.

Imagine a smart camera trap selectively clicking pictures to understand animal movement patterns within a particular habitat. These snapshots, or pieces of data captured from a data stream at adaptively chosen times, provide a glimpse of different animal movements unfolding through time. Learning a continuous-time process through snapshots, such as smart camera traps, is a central theme governing a wide array of online learning situations. In this paper, we adopt a learning-theoretic perspective in understanding the fundamental nature of learning different classes of functions from both discrete data streams and continuous data streams. In our first framework, the setting, a learning algorithm discretely queries from a process to update a predictor designed to make predictions given as input the data stream.