scholarly journals Compositionality Decomposed: How do Neural Networks Generalise?

2020 ◽  
Vol 67 ◽  
pp. 757-795
Author(s):  
Dieuwke Hupkes ◽  
Verna Dankers ◽  
Mathijs Mul ◽  
Elia Bruni
Keyword(s):  
Neural Networks ◽  
Compositional Data ◽  
Empirical Studies ◽  
Neural Model ◽  
Training Data ◽  
Data Set ◽  
Depth Analysis ◽  
The One ◽  
Transformer Model ◽  
Composition Operations

Despite a multitude of empirical studies, little consensus exists on whether neural networks are able to generalise compositionally, a controversy that, in part, stems from a lack of agreement about what it means for a neural model to be compositional. As a response to this controversy, we present a set of tests that provide a bridge between, on the one hand, the vast amount of linguistic and philosophical theory about compositionality of language and, on the other, the successful neural models of language. We collect different interpretations of compositionality and translate them into five theoretically grounded tests for models that are formulated on a task-independent level. In particular, we provide tests to investigate (i) if models systematically recombine known parts and rules (ii) if models can extend their predictions beyond the length they have seen in the training data (iii) if models’ composition operations are local or global (iv) if models’ predictions are robust to synonym substitutions and (v) if models favour rules or exceptions during training. To demonstrate the usefulness of this evaluation paradigm, we instantiate these five tests on a highly compositional data set which we dub PCFG SET and apply the resulting tests to three popular sequence-to-sequence models: a recurrent, a convolution-based and a transformer model. We provide an in-depth analysis of the results, which uncover the strengths and weaknesses of these three architectures and point to potential areas of improvement.

scholarly journals Compositionality Decomposed: How do Neural Networks Generalise? (Extended Abstract)

2020 ◽  
Author(s):  
Dieuwke Hupkes ◽  
Verna Dankers ◽  
Mathijs Mul ◽  
Elia Bruni
Keyword(s):  
Neural Networks ◽  
Compositional Data ◽  
Empirical Studies ◽  
The Other ◽  
Philosophical Theory ◽  
Neural Models ◽  
Data Set ◽  
Vast Amount ◽  
Depth Analysis ◽  
The One

Despite a multitude of empirical studies, little consensus exists on whether neural networks are able to generalise compositionally. As a response to this controversy, we present a set of tests that provide a bridge between, on the one hand, the vast amount of linguistic and philosophical theory about compositionality of language and, on the other, the successful neural models of language. We collect different interpretations of compositionality and translate them into five theoretically grounded tests for models that are formulated on a task-independent level. To demonstrate the usefulness of this evaluation paradigm, we instantiate these five tests on a highly compositional data set which we dub PCFG SET, apply the resulting tests to three popular sequence-to-sequence models and provide an in-depth analysis of the results.

scholarly journals Localization of Scattering Objects Using Neural Networks

Sensors ◽  
2020 ◽  
Vol 21 (1) ◽  
pp. 11
Author(s):  
Domonkos Haffner ◽  
Ferenc Izsák
Keyword(s):  
Neural Networks ◽  
Plane Waves ◽  
Measurement Data ◽  
Scattered Wave ◽  
Training Data ◽  
Data Set ◽  
Main Compound ◽  
Incident Plane ◽  
The Neural Networks ◽  
Simulation Package

The localization of multiple scattering objects is performed while using scattered waves. An up-to-date approach: neural networks are used to estimate the corresponding locations. In the scattering phenomenon under investigation, we assume known incident plane waves, fully reflecting balls with known diameters and measurement data of the scattered wave on one fixed segment. The training data are constructed while using the simulation package μ-diff in Matlab. The structure of the neural networks, which are widely used for similar purposes, is further developed. A complex locally connected layer is the main compound of the proposed setup. With this and an appropriate preprocessing of the training data set, the number of parameters can be kept at a relatively low level. As a result, using a relatively large training data set, the unknown locations of the objects can be estimated effectively.

scholarly journals Opportunities and Constraints in Applying Artificial Neural Networks (ANNs) in Food Authentication. Honey—A Case Study

2021 ◽  
Vol 11 (15) ◽  
pp. 6723
Author(s):  
Ariana Raluca Hategan ◽  
Romulus Puscas ◽  
Gabriela Cristea ◽  
Adriana Dehelean ◽  
Francois Guyon ◽  
...  
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Training Sample ◽  
Training Data ◽  
Elemental Content ◽  
Neural Structure ◽  
Food Authentication ◽  
Data Set ◽  
Artificial Neural

The present work aims to test the potential of the application of Artificial Neural Networks (ANNs) for food authentication. For this purpose, honey was chosen as the working matrix. The samples were originated from two countries: Romania (50) and France (53), having as floral origins: acacia, linden, honeydew, colza, galium verum, coriander, sunflower, thyme, raspberry, lavender and chestnut. The ANNs were built on the isotope and elemental content of the investigated honey samples. This approach conducted to the development of a prediction model for geographical recognition with an accuracy of 96%. Alongside this work, distinct models were developed and tested, with the aim of identifying the most suitable configurations for this application. In this regard, improvements have been continuously performed; the most important of them consisted in overcoming the unwanted phenomenon of over-fitting, observed for the training data set. This was achieved by identifying appropriate values for the number of iterations over the training data and for the size and number of the hidden layers and by introducing of a dropout layer in the configuration of the neural structure. As a conclusion, ANNs can be successfully applied in food authenticity control, but with a degree of caution with respect to the “over optimization” of the correct classification percentage for the training sample set, which can lead to an over-fitted model.

scholarly journals Keep it simple - A case study of model development in the context of the Dynamic Stocks and Flows (DSF) task

2010 ◽  
Vol 2 (2) ◽  
pp. 38-51 ◽  
Author(s):  
Marc Halbrügge
Keyword(s):  
Goodness Of Fit ◽  
Model Development ◽  
Cognitive Model ◽  
Training Data ◽  
Sequence Matching ◽  
Data Set ◽  
Depth Analysis ◽  
Stocks And Flows ◽  
Matching Techniques

Keep it simple - A case study of model development in the context of the Dynamic Stocks and Flows (DSF) taskThis paper describes the creation of a cognitive model submitted to the ‘Dynamic Stocks and Flows’ (DSF) modeling challenge. This challenge aims at comparing computational cognitive models for human behavior during an open ended control task. Participants in the modeling competition were provided with a simulation environment and training data for benchmarking their models while the actual specification of the competition task was withheld. To meet this challenge, the cognitive model described here was designed and optimized for generalizability. Only two simple assumptions about human problem solving were used to explain the empirical findings of the training data. In-depth analysis of the data set prior to the development of the model led to the dismissal of correlations or other parametric statistics as goodness-of-fit indicators. A new statistical measurement based on rank orders and sequence matching techniques is being proposed instead. This measurement, when being applied to the human sample, also identifies clusters of subjects that use different strategies for the task. The acceptability of the fits achieved by the model is verified using permutation tests.

Simple Convolutional-Based Models: Are They Learning the Task or the Data?

2021 ◽  
pp. 1-17
Author(s):  
Luis Sa-Couto ◽  
Andreas Wichert
Keyword(s):  
Neural Networks ◽  
Pattern Recognition ◽  
Training Data ◽  
Model Complexity ◽  
Data Sets ◽  
Simple Task ◽  
Data Set ◽  
Knowing That ◽  
Handwritten Digit ◽  
End To End

Abstract Convolutional neural networks (CNNs) evolved from Fukushima's neocognitron model, which is based on the ideas of Hubel and Wiesel about the early stages of the visual cortex. Unlike other branches of neocognitron-based models, the typical CNN is based on end-to-end supervised learning by backpropagation and removes the focus from built-in invariance mechanisms, using pooling not as a way to tolerate small shifts but as a regularization tool that decreases model complexity. These properties of end-to-end supervision and flexibility of structure allow the typical CNN to become highly tuned to the training data, leading to extremely high accuracies on typical visual pattern recognition data sets. However, in this work, we hypothesize that there is a flip side to this capability, a hidden overfitting. More concretely, a supervised, backpropagation based CNN will outperform a neocognitron/map transformation cascade (MTCCXC) when trained and tested inside the same data set. Yet if we take both models trained and test them on the same task but on another data set (without retraining), the overfitting appears. Other neocognitron descendants like the What-Where model go in a different direction. In these models, learning remains unsupervised, but more structure is added to capture invariance to typical changes. Knowing that, we further hypothesize that if we repeat the same experiments with this model, the lack of supervision may make it worse than the typical CNN inside the same data set, but the added structure will make it generalize even better to another one. To put our hypothesis to the test, we choose the simple task of handwritten digit classification and take two well-known data sets of it: MNIST and ETL-1. To try to make the two data sets as similar as possible, we experiment with several types of preprocessing. However, regardless of the type in question, the results align exactly with expectation.

scholarly journals Automatic Gully Detection: Neural Networks and Computer Vision

2020 ◽  
Vol 12 (11) ◽  
pp. 1743
Author(s):  
Artur M. Gafurov ◽  
Oleg P. Yermolayev
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Convolutional Neural Networks ◽  
Large Scale ◽  
Satellite Images ◽  
Gully Erosion ◽  
Training Data ◽  
Russian Plain ◽  
Data Set ◽  
High Resolution Satellite Images

Transition from manual (visual) interpretation to fully automated gully detection is an important task for quantitative assessment of modern gully erosion, especially when it comes to large mapping areas. Existing approaches to semi-automated gully detection are based on either object-oriented selection based on multispectral images or gully selection based on a probabilistic model obtained using digital elevation models (DEMs). These approaches cannot be used for the assessment of gully erosion on the territory of the European part of Russia most affected by gully erosion due to the lack of national large-scale DEM and limited resolution of open source multispectral satellite images. An approach based on the use of convolutional neural networks for automated gully detection on the RGB-synthesis of ultra-high resolution satellite images publicly available for the test region of the east of the Russian Plain with intensive basin erosion has been proposed and developed. The Keras library and U-Net architecture of convolutional neural networks were used for training. Preliminary results of application of the trained gully erosion convolutional neural network (GECNN) allow asserting that the algorithm performs well in detecting active gullies, well differentiates gullies from other linear forms of slope erosion — rills and balkas, but so far has errors in detecting complex gully systems. Also, GECNN does not identify a gully in 10% of cases and in another 10% of cases it identifies not a gully. To solve these problems, it is necessary to additionally train the neural network on the enlarged training data set.

scholarly journals Bokeh Effect Rendering with Vision Transformers

2022 ◽  
Author(s):  
Hariharan Nagasubramaniam ◽  
Rabih Younes
Keyword(s):  
Neural Networks ◽  
State Of The Art ◽  
Large Diameter ◽  
Depth Map ◽  
Global Information ◽  
Initial Training ◽  
Data Set ◽  
Current State ◽  
Mobile Cameras ◽  
Transformer Model

Bokeh effect is growing to be an important feature in photography, essentially to choose an object of interest to be in focus with the rest of the background being blurred. While naturally rendering this effect requires a DSLR with large diameter of aperture, with the current advancements in Deep Learning, this effect can also be produced in mobile cameras. Most of the existing methods use Convolutional Neural Networks while some relying on the depth map to render this effect. In this paper, we propose an end-to-end Vision Transformer model for Bokeh rendering of images from monocular camera. This architecture uses vision transformers as backbone, thus learning from the entire image rather than just the parts from the filters in a CNN. This property of retaining global information coupled with initial training of the model for image restoration before training to render the blur effect for the background, allows our method to produce clearer images and outperform the current state-of-the-art models on the EBB! Data set. The code to our proposed method can be found at: https://github.com/Soester10/ Bokeh-Rendering-with-Vision-Transformers.

Artificial neural networks in diagnosis of thyroid function from in vitro laboratory tests

1993 ◽  
Vol 39 (11) ◽  
pp. 2248-2253 ◽  
Author(s):  
P K Sharpe ◽  
H E Solberg ◽  
K Rootwelt ◽  
M Yearworth
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Thyroid Function ◽  
Laboratory Tests ◽  
Extreme Values ◽  
Real Life ◽  
Training Data ◽  
Data Set ◽  
Artificial Neural

Abstract We studied the potential benefit of using artificial neural networks (ANNs) for the diagnosis of thyroid function. We examined two types of ANN architecture and assessed their robustness in the face of diagnostic noise. The thyroid function data we used had previously been studied by multivariate statistical methods and a variety of pattern-recognition techniques. The total data set comprised 392 cases that had been classified according to both thyroid function and 19 clinical categories. All cases had a complete set of results of six laboratory tests (total thyroxine, free thyroxine, triiodothyronine, triiodothyronine uptake test, thyrotropin, and thyroxine-binding globulin). This data set was divided into subsets used for training the networks and for testing their performance; the test subsets contained various proportions of cases with diagnostic noise to mimic real-life diagnostic situations. The networks studied were a multilayer perceptron trained by back-propagation, and a learning vector quantization network. The training data subsets were selected according to two strategies: either training data based on cases with extreme values for the laboratory tests with randomly selected nonextreme cases added, or training cases from very pure functional groups. Both network architectures were efficient irrespective of the type of training data. The correct allocation of cases in test data subsets was 96.4-99.7% when extreme values were used for training and 92.7-98.8% when only pure cases were used.

Data set entity recognition based on distant supervision

2021 ◽  
Vol ahead-of-print (ahead-of-print) ◽  
Author(s):  
Pengcheng Li ◽  
Qikai Liu ◽  
Qikai Cheng ◽  
Wei Lu
Keyword(s):  
Supervised Learning ◽  
Large Scale ◽  
Data Augmentation ◽  
Scientific Literature ◽  
Neural Model ◽  
Training Data ◽  
Entity Recognition ◽  
Data Set ◽  
Content Type ◽  
Augmentation Techniques

Purpose This paper aims to identify data set entities in scientific literature. To address poor recognition caused by a lack of training corpora in existing studies, a distant supervised learning-based approach is proposed to identify data set entities automatically from large-scale scientific literature in an open domain. Design/methodology/approach Firstly, the authors use a dictionary combined with a bootstrapping strategy to create a labelled corpus to apply supervised learning. Secondly, a bidirectional encoder representation from transformers (BERT)-based neural model was applied to identify data set entities in the scientific literature automatically. Finally, two data augmentation techniques, entity replacement and entity masking, were introduced to enhance the model generalisability and improve the recognition of data set entities. Findings In the absence of training data, the proposed method can effectively identify data set entities in large-scale scientific papers. The BERT-based vectorised representation and data augmentation techniques enable significant improvements in the generality and robustness of named entity recognition models, especially in long-tailed data set entity recognition. Originality/value This paper provides a practical research method for automatically recognising data set entities in scientific literature. To the best of the authors’ knowledge, this is the first attempt to apply distant learning to the study of data set entity recognition. The authors introduce a robust vectorised representation and two data augmentation strategies (entity replacement and entity masking) to address the problem inherent in distant supervised learning methods, which the existing research has mostly ignored. The experimental results demonstrate that our approach effectively improves the recognition of data set entities, especially long-tailed data set entities.

Sign in / Sign up

Export Citation Format

Share Document