Few-shot learning refers to understanding new concepts from only a few examples. We propose an information retrieval-inspired approach for this problem that is motivated by the increased importance of maximally leveraging all the available information in this low-data regime. We define a training objective that aims to extract as much information as possible from each training batch by effectively optimizing over all relative orderings of the batch points simultaneously. In particular, we view each batch point as a query' that ranks the remaining ones based on its predicted relevance to them and we define a model within the framework of structured prediction to optimize mean Average Precision over these rankings. Our method achieves impressive results on the standard few-shot classification benchmarks while is also capable of few-shot retrieval.

The accuracy of information retrieval systems is often measured using average precision (AP). Given a set of positive (relevant) and negative (non-relevant) samples, the parameters of a retrieval system can be estimated using the AP-SVM framework, which minimizes a regularized convex upper bound on the empirical AP loss. However, the high computational complexity of loss-augmented inference, which is required for learning an AP-SVM, prohibits its use with large training datasets. To alleviate this deficiency, we propose three complementary approaches. The second approach takes advantage of the fact that we do not require a full ranking during loss-augmented inference.

Most generative models for clustering implicitly assume that the number of data points in each cluster grows linearly with the total number of data points. Finite mixture models, Dirichlet process mixture models, and Pitman--Yor process mixture models make this assumption, as do all other infinitely exchangeable clustering models. However, for some applications, this assumption is inappropriate. For example, when performing entity resolution, the size of each cluster should be unrelated to the size of the data set, and each cluster should contain a negligible fraction of the total number of data points. These applications require models that yield clusters whose sizes grow sublinearly with the size of the data set.

Binary hashing is a well-known approach for fast approximate nearest-neighbor search in information retrieval. Much work has focused on affinity-based objective functions involving the hash functions or binary codes. These objective functions encode neighborhood information between data points and are often inspired by manifold learning algorithms. They ensure that the hash functions differ from each other through constraints or penalty terms that encourage codes to be orthogonal or dissimilar across bits, but this couples the binary variables and complicates the already difficult optimization. We propose a much simpler approach: we train each hash function (or bit) independently from each other, but introduce diversity among them using techniques from classifier ensembles.

Click-through data has proven to be a valuable resource for improving search-ranking quality. Search engines can easily collect click data, but biases introduced in the data can make it difficult to use the data effectively. In order to measure the effects of biases, many click models have been proposed in the literature. However, none of the models can explain the observation that users with different search intent (e.g., informational, navigational, etc.) have different click behaviors. In this paper, we study how differences in user search intent can influence click activities and determined that there exists a bias between user search intent and the relevance of the document relevance. Based on this observation, we propose a search intent bias hypothesis that can be applied to most existing click models to improve their ability to learn unbiased relevance. Experimental results demonstrate that after adopting the search intent hypothesis, click models can better interpret user clicks and substantially improve retrieval performance.

In machine learning, the notion of multi-armed bandits (MAB) refers to a class of online learning problems, in which a learner is supposed to simultaneously explore and exploit a given set of choice alternatives (metaphorically referred to as "arms") in the course of a sequential decision process (Lattimore and Szepesvári, 2019). In this paper, we consider an extension of the basic setting, which is practically motivated by the problem of preselection as recently introduced by Saha and Gopalan (2018b) and Bengs and Hüllermeier (2019): Instead of selecting a single arm, the learner is only supposed to preselect a promising subset of arms. The final choice is then made by a selector, for example a human user or another algorithm. In information retrieval, for instance, the role of the learner is played by a search engine, and the selector is the user who seeks a certain information. Another application, which served as a concrete motivation of our setting and will also be used in our experimental study, is the problem of algorithm (pre-)selection (Kerschke et al., 2018).

We consider the large-scale query-document retrieval problem: given a query (e.g., a question), return the set of relevant documents (e.g., paragraphs containing the answer) from a large document corpus. This problem is often solved in two steps. The retrieval phase first reduces the solution space, returning a subset of candidate documents. The scoring phase then re-ranks the documents. Critically, the retrieval algorithm not only desires high recall but also requires to be highly efficient, returning candidates in time sublinear to the number of documents. Unlike the scoring phase witnessing significant advances recently due to the BERT-style pre-training tasks on cross-attention models, the retrieval phase remains less well studied. Most previous works rely on classic Information Retrieval (IR) methods such as BM-25 (token matching + TF-IDF weights). These models only accept sparse handcrafted features and can not be optimized for different downstream tasks of interest. In this paper, we conduct a comprehensive study on the embedding-based retrieval models. We show that the key ingredient of learning a strong embedding-based Transformer model is the set of pre-training tasks. With adequately designed paragraph-level pre-training tasks, the Transformer models can remarkably improve over the widely-used BM-25 as well as embedding models without Transformers. The paragraph-level pre-training tasks we studied are Inverse Cloze Task (ICT), Body First Selection (BFS), Wiki Link Prediction (WLP), and the combination of all three.

Knowledge graphs have been shown to play an important role in recent knowledge mining and discovery, for example in the field of life sciences or bioinformatics. Although a lot of research has been done on the field of query optimization, query transformation and of course in storing and retrieving large scale knowledge graphs the field of algorithmic optimization is still a major challenge and a vital factor in using graph databases. Few researchers have addressed the problem of optimizing algorithms on large scale labeled property graphs. Here, we present two optimization approaches and compare them with a naive approach of directly querying the graph database. The aim of our work is to determine limiting factors of graph databases like Neo4j and we describe a novel solution to tackle these challenges. For this, we suggest a classification schema to differ between the complexity of a problem on a graph database. We evaluate our optimization approaches on a test system containing a knowledge graph derived biomedical publication data enriched with text mining data. This dense graph has more than 71M nodes and 850M relationships. The results are very encouraging and - depending on the problem - we were able to show a speedup of a factor between 44 and 3839.

You need either "Information" or "Knowledge". HUNT from Vistaar is "AI driven" search of our internal repository consisting of over 980K regulations and guidance documents from global health authorities which gets updated daily. So get your "Authoritative" documents from different countries at 1 place and get them in 1 search. Be sure about the documents.

Logic-based systems for query answering over knowledge graphs return only answers that rely on information explicitly represented in the graph. To improve recall, recent works have proposed the use of embeddings to predict additional information like missing links, or labels. These embeddings enable scoring entities in the graph as the answer a query, without being fully dependent on the graph structure. In its simplest case, answering a query in such a setting requires predicting a link between two entities. However, link prediction is not sufficient to address complex queries that involve multiple entities and variables. To solve this task, we propose to apply a message passing mechanism to a graph representation of the query, where nodes correspond to variables and entities. This results in an embedding of the query, such that answering entities are close to it in the embedding space. The general formulation of our method allows it to encode a more diverse set of query types in comparison to previous work. We evaluate our method by answering queries that rely on edges not seen during training, obtaining competitive performance. In contrast with previous work, we show that our method can generalize from training for the single-hop, link prediction task, to answering queries with more complex structures. A qualitative analysis reveals that the learned embeddings successfully capture the notion of different entity types.