ICPR2020 Paper Browser

Paper download is intended for registered attendees only, and is subjected to the IEEE Copyright Policy. Any other use is strongly forbidden.

An Efficient Empirical Solver for Localized Multiple Kernel Learning Via DNNs

Ziming Zhang

Auto-TLDR; Localized Multiple Kernel Learning using LMKL-Net

Abstract Slides Poster

In this paper we propose solving localized multiple kernel learning (LMKL) using LMKL-Net, a feedforward deep neural network (DNN). In contrast to previous works, as a learning principle we propose parameterizing the gating function for learning kernel combination weights and the multiclass classifier using an attentional network (AN) and a multilayer perceptron (MLP), respectively. Such interpretability helps us better understand how the network solves the problem. Thanks to stochastic gradient descent (SGD), our approach has {\em linear} computational complexity in training. Empirically on benchmark datasets we demonstrate that with comparable or better accuracy than the state-of-the-art, our LMKL-Net can be trained about {\bf two orders of magnitude} faster with about {\bf two orders of magnitude} smaller memory footprint for large-scale learning.

Similar papers

RNN Training along Locally Optimal Trajectories via Frank-Wolfe Algorithm

Yun Yue, Ming Li, Venkatesh Saligrama, Ziming Zhang

Auto-TLDR; Frank-Wolfe Algorithm for Efficient Training of RNNs

Abstract Slides Poster Similar

We propose a novel and efficient training method for RNNs by iteratively seeking a local minima on the loss surface within a small region, and leverage this directional vector for the update, in an outer-loop. We propose to utilize the Frank-Wolfe (FW) algorithm in this context. Although, FW implicitly involves normalized gradients, which can lead to a slow convergence rate, we develop a novel RNN training method that, surprisingly, even with the additional cost, the overall training cost is empirically observed to be lower than back-propagation. Our method leads to a new Frank-Wolfe method, that is in essence an SGD algorithm with a restart scheme. We prove that under certain conditions our algorithm has a sublinear convergence rate of $O(1/\epsilon)$ for $\epsilon$ error. We then conduct empirical experiments on several benchmark datasets including those that exhibit long-term dependencies, and show significant performance improvement. We also experiment with deep RNN architectures and show efficient training performance. Finally, we demonstrate that our training method is robust to noisy data.

Learning Sparse Deep Neural Networks Using Efficient Structured Projections on Convex Constraints for Green AI

Michel Barlaud, Frederic Guyard

Auto-TLDR; Constrained Deep Neural Network with Constrained Splitting Projection

Abstract Slides Poster Similar

In recent years, deep neural networks (DNN) have been applied to different domains and achieved dramatic performance improvements over state-of-the-art classical methods. These performances of DNNs were however often obtained with networks containing millions of parameters and which training required heavy computational power. In order to cope with this computational issue a huge literature deals with proximal regularization methods which are time consuming.\\ In this paper, we propose instead a constrained approach. We provide the general framework for our new splitting projection gradient method. Our splitting algorithm iterates a gradient step and a projection on convex sets. We study algorithms for different constraints: the classical $\ell_1$ unstructured constraint and structured constraints such as the nuclear norm, the $\ell_{2,1} $ constraint (Group LASSO). We propose a new $\ell_{1,1} $ structured constraint for which we provide a new projection algorithm We demonstrate the effectiveness of our method on three popular datasets (MNIST, Fashion MNIST and CIFAR). Experiments on these datasets show that our splitting projection method with our new $\ell_{1,1} $ structured constraint provides the best reduction of memory and computational power. Experiments show that fully connected linear DNN are more efficient for green AI.

Low-Cost Lipschitz-Independent Adaptive Importance Sampling of Stochastic Gradients

Huikang Liu, Xiaolu Wang, Jiajin Li, Man-Cho Anthony So

Auto-TLDR; Adaptive Importance Sampling for Stochastic Gradient Descent

Abstract Slides Similar

Stochastic gradient descent (SGD) usually samples training data based on the uniform distribution, which may not be a good choice because of the high variance of its stochastic gradient. Thus, importance sampling methods are considered in the literature to improve the performance. Most previous work on SGD-based methods with importance sampling requires the knowledge of Lipschitz constants of all component gradients, which are in general difficult to estimate. In this paper, we study an adaptive importance sampling method for common SGD-based methods by exploiting the local first-order information without knowing any Lipschitz constants. In particular, we periodically changes the sampling distribution by only utilizing the gradient norms in the past few iterations. We prove that our adaptive importance sampling non-asymptotically reduces the variance of the stochastic gradients in SGD, and thus better convergence bounds than that for vanilla SGD can be obtained. We extend this sampling method to several other widely used stochastic gradient algorithms including SGD with momentum and ADAM. Experiments on common convex learning problems and deep neural networks illustrate notably enhanced performance using the adaptive sampling strategy.

Mean Decision Rules Method with Smart Sampling for Fast Large-Scale Binary SVM Classification

Alexandra Makarova, Mikhail Kurbakov, Valentina Sulimova

Auto-TLDR; Improving Mean Decision Rule for Large-Scale Binary SVM Problems

Abstract Slides Poster Similar

This paper relies on the Mean Decision Rule (MDR) method for solving large-scale binary SVM problems. It consists in taking small random samples of the full dataset and separate training for each of them with consecutive averaging the respective individual decision rules to obtain a final one. This paper proposes two new approaches to improve it. The first proposed approach is a new sampling technique that exploits SVM and MDR properties to fast form so called smart samples by selecting only the objects, that are candidates to be the support ones. The proposed technique essentially increases MDR convergence and allows to reach the highest quality in less time. In the case of kernel-based MDR (KMDR) the proposed sampling technique allows additionally to reduce the number of support objects in the final decision rule and, as a result, to decrease the recognition time. The second proposed approach is a new data strategy to accelerate random access to large datasets stored in the traditional libsvm format. The proposed strategy allows to quickly extract random subsets of objects from a file and load them into RAM, and is it also suitable for any sampling-based methods, including stochastic gradient methods. Joint using of the proposed approaches with (K)MDR allows to obtain the best (or near the best) decision of large-scale binary SVM problems faster, compared to the existing SVM solvers.

Norm Loss: An Efficient yet Effective Regularization Method for Deep Neural Networks

Theodoros Georgiou, Sebastian Schmitt, Thomas Baeck, Wei Chen, Michael Lew

Auto-TLDR; Weight Soft-Regularization with Oblique Manifold for Convolutional Neural Network Training

Abstract Slides Poster Similar

Convolutional neural network training can suffer from diverse issues like exploding or vanishing gradients, scaling-based weight space symmetry and covariant-shift. In order to address these issues, researchers develop weight regularization methods and activation normalization methods. In this work we propose a weight soft-regularization method based on the Oblique manifold. The proposed method uses a loss function which pushes each weight vector to have a norm close to one, i.e. the weight matrix is smoothly steered toward the so-called Oblique manifold. We evaluate our method on the very popular CIFAR-10, CIFAR-100 and ImageNet 2012 datasets using two state-of-the-art architectures, namely the ResNet and wide-ResNet. Our method introduces negligible computational overhead and the results show that it is competitive to the state-of-the-art and in some cases superior to it. Additionally, the results are less sensitive to hyperparameter settings such as batch size and regularization factor.

Meta Generalized Network for Few-Shot Classification

Wei Wu, Shanmin Pang, Zhiqiang Tian, Yaochen Li

Auto-TLDR; Meta Generalized Network for Few-Shot Classification

Abstract Similar

Few-shot classification aims to learn a well performance model with very limited labeled examples. There are mainly two directions for this aim, namely, meta- and metric-learning. Meta learning trains models in a particular way to fast adapt to new tasks, but it neglects variational features of images. Metric learning considers relationships among same or different classes, however on the downside, it usually fails to achieve competitive performance on unseen boundary examples. In this paper, we propose a Meta Generalized Network (MGNet) that aims to combine advantages of both meta- and metric-learning. There are two novel components in MGNet. Specifically, we first develop a meta backbone training method that both learns a flexible feature extractor and a classifier initializer efficiently, delightedly leading to fast adaption to unseen few-shot tasks without overfitting. Second, we design a trainable adaptive interval model to improve the cosine classifier, which increases the recognition accuracy of hard examples. We train the meta backbone in the training stage by all classes, and fine-tune the meta-backbone as well as train the adaptive classifier in the testing stage.

Improving Batch Normalization with Skewness Reduction for Deep Neural Networks

Pak Lun Kevin Ding, Martin Sarah, Baoxin Li

Auto-TLDR; Batch Normalization with Skewness Reduction

Abstract Slides Poster Similar

Batch Normalization (BN) is a well-known technique used in training deep neural networks. The main idea behind batch normalization is to normalize the features of the layers ($i.e.$, transforming them to have a mean equal to zero and a variance equal to one). Such a procedure encourages the optimization landscape of the loss function to be smoother, and improve the learning of the networks for both speed and performance. In this paper, we demonstrate that the performance of the network can be improved, if the distributions of the features of the output in the same layer are similar. As normalizing based on mean and variance does not necessarily make the features to have the same distribution, we propose a new normalization scheme: Batch Normalization with Skewness Reduction (BNSR). Comparing with other normalization approaches, BNSR transforms not just only the mean and variance, but also the skewness of the data. By tackling this property of a distribution, we are able to make the output distributions of the layers to be further similar. The nonlinearity of BNSR may further improve the expressiveness of the underlying network. Comparisons with other normalization schemes are tested on the CIFAR-100 and ImageNet datasets. Experimental results show that the proposed approach can outperform other state-of-the-arts that are not equipped with BNSR.

Learning Sign-Constrained Support Vector Machines

Kenya Tajima, Kouhei Tsuchida, Esmeraldo Ronnie Rey Zara, Naoya Ohta, Tsuyoshi Kato

Auto-TLDR; Constrained Sign Constraints for Learning Linear Support Vector Machine

Abstract Poster Similar

Domain knowledge is useful to improve the generalization performance of learning machines. Sign constraints are a handy representation to combine domain knowledge with learning machine. In this paper, we consider constraining the signs of the weight coefficients in learning the linear support vector machine, and develop two optimization algorithms for minimizing the empirical risk under the sign constraints. One of the two algorithms is based on the projected gradient method, in which each iteration of the projected gradient method takes O(nd) computational cost and the sublinear convergence of the objective error is guaranteed. The second algorithm is based on the Frank-Wolfe method that also converges sublinearly and possesses a clear termination criterion. We show that each iteration of the Frank-Wolfe also requires O(nd) cost. Furthermore, we derive the explicit expression for the minimal iteration number to ensure an epsilon-accurate solution by analyzing the curvature of the objective function. Finally, we empirically demonstrate that the sign constraints are a promising technique when similarities to the training examples compose the feature vector.

Is the Meta-Learning Idea Able to Improve the Generalization of Deep Neural Networks on the Standard Supervised Learning?

Xiang Deng, Zhongfei Zhang

Auto-TLDR; Meta-learning Based Training of Deep Neural Networks for Few-Shot Learning

Abstract Slides Poster Similar

Substantial efforts have been made on improving the generalization abilities of deep neural networks (DNNs) in order to obtain better performances without introducing more parameters. On the other hand, meta-learning approaches exhibit powerful generalization on new tasks in few-shot learning. Intuitively, few-shot learning is more challenging than the standard supervised learning as each target class only has a very few or no training samples. The natural question that arises is whether the meta-learning idea can be used for improving the generalization of DNNs on the standard supervised learning. In this paper, we propose a novel meta-learning based training procedure (MLTP) for DNNs and demonstrate that the meta-learning idea can indeed improve the generalization abilities of DNNs. MLTP simulates the meta-training process by considering a batch of training samples as a task. The key idea is that the gradient descent step for improving the current task performance should also improve a new task performance, which is ignored by the current standard procedure for training neural networks. MLTP also benefits from all the existing training techniques such as dropout, weight decay, and batch normalization. We evaluate MLTP by training a variety of small and large neural networks on three benchmark datasets, i.e., CIFAR-10, CIFAR-100, and Tiny ImageNet. The experimental results show a consistently improved generalization performance on all the DNNs with different sizes, which verifies the promise of MLTP and demonstrates that the meta-learning idea is indeed able to improve the generalization of DNNs on the standard supervised learning.

Generalization Comparison of Deep Neural Networks Via Output Sensitivity

Mahsa Forouzesh, Farnood Salehi, Patrick Thiran

Auto-TLDR; Generalization of Deep Neural Networks using Sensitivity

Abstract Slides Similar

Although recent works have brought some insights into the performance improvement of techniques used in state-of-the-art deep-learning models, more work is needed to understand their generalization properties. We shed light on this matter by linking the loss function to the output's sensitivity to its input. We find a rather strong empirical relation between the output sensitivity and the variance in the bias-variance decomposition of the loss function, which hints on using sensitivity as a metric for comparing the generalization performance of networks, without requiring labeled data. We find that sensitivity is decreased by applying popular methods which improve the generalization performance of the model, such as (1) using a deep network rather than a wide one, (2) adding convolutional layers to baseline classifiers instead of adding fully-connected layers, (3) using batch normalization, dropout and max-pooling, and (4) applying parameter initialization techniques.

Unveiling Groups of Related Tasks in Multi-Task Learning

Jordan Frecon, Saverio Salzo, Massimiliano Pontil

Auto-TLDR; Continuous Bilevel Optimization for Multi-Task Learning

Abstract Slides Poster Similar

A common approach in multi-task learning is to encourage the tasks to share a low dimensional representation. This has led to the popular method of trace norm regularization, which has proved effective in many applications. In this paper, we extend this approach by allowing the tasks to partition into different groups, within which trace norm regularization is separately applied. We propose a continuous bilevel optimization framework to simultaneously identify groups of related tasks and learn a low dimensional representation within each group. Hinging on recent results on the derivative of generalized matrix functions, we devise a smooth approximation of the upper-level objective via a dual forward-backward algorithm with Bregman distances. This allows us to solve the bilevel problem by a gradient-based scheme. Numerical experiments on synthetic and benchmark datasets support the effectiveness of the proposed method.

Learning Stable Deep Predictive Coding Networks with Weight Norm Supervision

Guo Ruohao

Auto-TLDR; Stability of Predictive Coding Network with Weight Norm Supervision

Abstract Slides Poster Similar

Predictive Coding Network (PCN) is an important neural network inspired by visual processing models in neuroscience. It combines the feedforward and feedback processing and has the architecture of recurrent neural networks (RNNs). This type of network is usually trained with backpropagation through time (BPTT). With infinite recurrent steps, PCN is a dynamic system. However, as one of the most important properties, stability is rarely studied in this type of network. Inspired by reservoir computing, we investigate the stability of hierarchical RNNs from the perspective of dynamic systems, and propose a sufficient condition for their echo state property (ESP). Our study shows the global stability is determined by stability of the local layers and the feedback between neighboring layers. Based on it, we further propose Weight Norm Supervision, a new algorithm that controls the stability of PCN dynamics by imposing different weight norm constraints on different parts of the network. We compare our approach with other training methods in terms of stability and prediction capability. The experiments show that our algorithm learns stable PCNs with a reliable prediction precision in the most effective and controllable way.

3CS Algorithm for Efficient Gaussian Process Model Retrieval

Fabian Berns, Kjeld Schmidt, Ingolf Bracht, Christian Beecks

Auto-TLDR; Efficient retrieval of Gaussian Process Models for large-scale data using divide-&-conquer-based approach

Abstract Slides Poster Similar

Gaussian Process Models (GPMs) have been applied for various pattern recognition tasks due to their analytical tractability, ability to quantify uncertainty for their own results as well as to subsume prominent other regression techniques. Despite these promising prospects their super-quadratic computation time complexity for model selection and evaluation impedes its broader application for more than a few thousand data points. Although there have been many proposals towards Gaussian Processes for large-scale data, those only offer a linearly scaling improvement to a cubical scaling problem. In particular, solutions like the Nystrom approximation or sparse matrices are only taking fractions of the given data into account and subsequently lead to inaccurate models. In this paper, we thus propose a divide-&-conquer-based approach, that allows to efficiently retrieve GPMs for large-scale data. The resulting model is composed of independent pattern representations for non-overlapping segments of the given data and consequently reduces computation time significantly. Our performance analysis indicates that our proposal is able to outperform state-of-the-art algorithms for GPM retrieval with respect to the qualities of efficiency and accuracy.

E-DNAS: Differentiable Neural Architecture Search for Embedded Systems

Javier García López, Antonio Agudo, Francesc Moreno-Noguer

Auto-TLDR; E-DNAS: Differentiable Architecture Search for Light-Weight Networks for Image Classification

Abstract Slides Poster Similar

Designing optimal and light weight networks to fit in resource-limited platforms like mobiles, DSPs or GPUs is a challenging problem with a wide range of interesting applications, {\em e.g.} in embedded systems for autonomous driving. While most approaches are based on manual hyperparameter tuning, there exist a new line of research, the so-called NAS (Neural Architecture Search) methods, that aim to optimize several metrics during the design process, including memory requirements of the network, number of FLOPs, number of MACs (Multiply-ACcumulate operations) or inference latency. However, while NAS methods have shown very promising results, they are still significantly time and cost consuming. In this work we introduce E-DNAS, a differentiable architecture search method, which improves the efficiency of NAS methods in designing light-weight networks for the task of image classification. Concretely, E-DNAS computes, in a differentiable manner, the optimal size of a number of meta-kernels that capture patterns of the input data at different resolutions. We also leverage on the additive property of convolution operations to merge several kernels with different compatible sizes into a single one, reducing thus the number of operations and the time required to estimate the optimal configuration. We evaluate our approach on several datasets to perform classification. We report results in terms of the SoC (System on Chips) metric, typically used in the Texas Instruments TDA2x families for autonomous driving applications. The results show that our approach allows designing low latency architectures significantly faster than state-of-the-art.

Killing Four Birds with One Gaussian Process: The Relation between Different Test-Time Attacks

Kathrin Grosse, Michael Thomas Smith, Michael Backes

Auto-TLDR; Security of Gaussian Process Classifiers against Attack Algorithms

Abstract Slides Poster Similar

In machine learning (ML) security, attacks like evasion, model stealing or membership inference are generally studied in individually. Previous work has also shown a relationship between some attacks and decision function curvature of the targeted model. Consequently, we study an ML model allowing direct control over the decision surface curvature: Gaussian Process classifiers (GPCs). For evasion, we find that changing GPC's curvature to be robust against one attack algorithm boils down to enabling a different norm or attack algorithm to succeed. This is backed up by our formal analysis showing that static security guarantees are opposed to learning. Concerning intellectual property, we show formally that lazy learning does not necessarily leak all information when applied. In practice, often a seemingly secure curvature can be found. For example, we are able to secure GPC against empirical membership inference by proper configuration. In this configuration, however, the GPC's hyper-parameters are leaked, e.g. model reverse engineering succeeds. We conclude that attacks on classification should not be studied in isolation, but in relation to each other.

On Resource-Efficient Bayesian Network Classifiers and Deep Neural Networks

Wolfgang Roth, Günther Schindler, Holger Fröning, Franz Pernkopf

Auto-TLDR; Quantization-Aware Bayesian Network Classifiers for Small-Scale Scenarios

Abstract Slides Poster Similar

We present two methods to reduce the complexity of Bayesian network (BN) classifiers. First, we introduce quantization-aware training using the straight-through gradient estimator to quantize the parameters of BNs to few bits. Second, we extend a recently proposed differentiable tree-augmented naive Bayes (TAN) structure learning approach to also consider the model size. Both methods are motivated by recent developments in the deep learning community, and they provide effective means to trade off between model size and prediction accuracy, which is demonstrated in extensive experiments. Furthermore, we contrast quantized BN classifiers with quantized deep neural networks (DNNs) for small-scale scenarios which have hardly been investigated in the literature. We show Pareto optimal models with respect to model size, number of operations, and test error and find that both model classes are viable options.

Classification and Feature Selection Using a Primal-Dual Method and Projections on Structured Constraints

Michel Barlaud, Antonin Chambolle, Jean_Baptiste Caillau

Auto-TLDR; A Constrained Primal-dual Method for Structured Feature Selection on High Dimensional Data

Abstract Slides Poster Similar

This paper deals with feature selection using supervised classification on high dimensional datasets. A classical approach is to project data on a low dimensional space and classify by minimizing an appropriate quadratic cost. Our first contribution is to introduce a matrix of centers in the definition of this cost. Moreover, as quadratic costs are not robust to outliers, we propose to use an $\ell_1$ cost instead (or Huber loss to mitigate overfitting issues). While control on sparsity is commonly obtained by adding an $\ell_1$ constraint on the vectorized matrix of weights used for projecting the data, our second contribution is to enforce structured sparsity. To this end we propose constraints that take into account the matrix structure of the data, based either on the nuclear norm, on the $\ell_{2,1}$ norm, or on the $\ell_{1,2}$ norm for which we provide a new projection algorithm. We optimize simultaneously the projection matrix and the matrix of centers thanks to a new tailored constrained primal-dual method. The primal-dual framework is general enough to encompass the various robust losses and structured constraints we use, and allows a convergence analysis. We demonstrate the effectiveness of the approach on three biological datasets. Our primal-dual method with robust losses, adaptive centers and structured constraints does significantly better than classical methods, both in terms of accuracy and computational time.

Automatically Mining Relevant Variable Interactions Via Sparse Bayesian Learning

Ryoichiro Yafune, Daisuke Sakuma, Yasuo Tabei, Noritaka Saito, Hiroto Saigo

Auto-TLDR; Sparse Bayes for Interpretable Non-linear Prediction

Abstract Slides Poster Similar

With the rapid increase in the availability of large amount of data, prediction is becoming increasingly popular, and has widespread through our daily life. However, powerful non- linear prediction methods such as deep learning and SVM suffer from interpretability problem, making it hard to use in domains where the reason for decision making is required. In this paper, we develop an interpretable non-linear model called itemset Sparse Bayes (iSB), which builds a Bayesian probabilistic model, while simultaneously considering variable interactions. In order to suppress the resulting large number of variables, sparsity is imposed on regression weights by a sparsity inducing prior. As a subroutine to search for variable interactions, itemset enumeration algorithm is employed with a novel bounding condition. In computational experiments using real-world dataset, the proposed method performed better than decision tree by 10% in terms of r-squared . We also demonstrated the advantage of our method in Bayesian optimization setting, in which the proposed approach could successfully find the maximum of an unknown function faster than Gaussian process. The interpretability of iSB is naturally inherited to Bayesian optimization, thereby gives us a clue to understand which variables interactions are important in optimizing an unknown function.

Nearest Neighbor Classification Based on Activation Space of Convolutional Neural Network

Xinbo Ju, Shuo Shao, Huan Long, Weizhe Wang

Auto-TLDR; Convolutional Neural Network with Convex Hull Based Classifier

Abstract Poster Similar

In this paper, we propose a new image classifier based on the incorporation of the nearest neighbor algorithm and the activation space of convolutional neural network. The classifier has been successfully used on some state-of-the-art models and further improve their performance. Main technique tools we used are convex hull based classification and its acceleration. We find that 1) in several cases, the classifier can reach higher accuracy than original CNN; 2) by sampling, the classifier can work more efficiently; 3) centroid of each convex hull shows surprising ability in classification. Most of the work has strong geometry meanings, which helps us have a new understanding about convolutional layers.

A New Convex Loss Function for Multiple Instance Support Vector Machines

Sang-Baeg Kim, Jung-Man Bae

Auto-TLDR; WR-SVM: A Novel Multiple Instance SVM for Video Classification

Abstract Slides Poster Similar

We develop a novel multiple instance SVM based on maximizing the minimum witness rate (WR) among positive bags. We solve this nonlinear integer programming problem by the multiple instance SVM formulation with a convex loss function, where unknown integer labels of instances in positive bags are relaxed to real numbers between -1 and 1 using the tanh(.) to estimate WR of a positive bag. Our new model, WR-SVM, also can incorporates the multiple instance learning (MIL) problem into a simple deep neural network framework with no additional MIL pooling layers. WR-SVM is robust to input perturbation by eliminating the imbalance between positive instances allocated to positive bags. A better generalization ability is expected by the large margin due to the balanced allocation of positive instances to positive bags. We perform experiments on various video datasets to verify the effectiveness of our method for video classification. The results of WR-SVM outperform the state-of-the-art for MIL-based video classification models.

Learning with Multiplicative Perturbations

Xiulong Yang, Shihao Ji

Auto-TLDR; XAT and xVAT: A Multiplicative Adversarial Training Algorithm for Robust DNN Training

Abstract Slides Poster Similar

Adversarial Training (AT) and Virtual Adversarial Training (VAT) are the regularization techniques that train Deep Neural Networks (DNNs) with adversarial examples generated by adding small but worst-case perturbations to input examples. In this paper, we propose xAT and xVAT, new adversarial training algorithms that generate multiplicative perturbations to input examples for robust training of DNNs. Such perturbations are much more perceptible and interpretable than their additive counterparts exploited by AT and VAT. Furthermore, the multiplicative perturbations can be generated transductively or inductively, while the standard AT and VAT only support a transductive implementation. We conduct a series of experiments that analyze the behavior of the multiplicative perturbations and demonstrate that xAT and xVAT match or outperform state-of-the-art classification accuracies across multiple established benchmarks while being about 30% faster than their additive counterparts. Our source code can be found at https://github.com/sndnyang/xvat

Separation of Aleatoric and Epistemic Uncertainty in Deterministic Deep Neural Networks

Denis Huseljic, Bernhard Sick, Marek Herde, Daniel Kottke

Auto-TLDR; AE-DNN: Modeling Uncertainty in Deep Neural Networks

Abstract Slides Poster Similar

Despite the success of deep neural networks (DNN) in many applications, their ability to model uncertainty is still significantly limited. For example, in safety-critical applications such as autonomous driving, it is crucial to obtain a prediction that reflects different types of uncertainty to address life-threatening situations appropriately. In such cases, it is essential to be aware of the risk (i.e., aleatoric uncertainty) and the reliability (i.e., epistemic uncertainty) that comes with a prediction. We present AE-DNN, a model allowing the separation of aleatoric and epistemic uncertainty while maintaining a proper generalization capability. AE-DNN is based on deterministic DNN, which can determine the respective uncertainty measures in a single forward pass. In analyses with synthetic and image data, we show that our method improves the modeling of epistemic uncertainty while providing an intuitively understandable separation of risk and reliability.

Image Representation Learning by Transformation Regression

Xifeng Guo, Jiyuan Liu, Sihang Zhou, En Zhu, Shihao Dong

Auto-TLDR; Self-supervised Image Representation Learning using Continuous Parameter Prediction

Abstract Slides Poster Similar

Self-supervised learning is a thriving research direction since it can relieve the burden of human labeling for machine learning by seeking for supervision from data instead of human annotation. Although demonstrating promising performance in various applications, we observe that the existing methods usually model the auxiliary learning tasks as classification tasks with finite discrete labels, leading to insufficient supervisory signals, which in turn restricts the representation quality. In this paper, to solve the above problem and make full use of the supervision from data, we design a regression model to predict the continuous parameters of a group of transformations, i.e., image rotation, translation, and scaling. Surprisingly, this naive modification stimulates tremendous potential from data and the resulting supervisory signal has largely improved the performance of image representation learning. Extensive experiments on four image datasets, including CIFAR10, CIFAR100, STL10, and SVHN, indicate that our proposed algorithm outperforms the state-of-the-art unsupervised learning methods by a large margin in terms of classification accuracy. Crucially, we find that with our proposed training mechanism as an initialization, the performance of the existing state-of-the-art classification deep architectures can be preferably improved.

A Close Look at Deep Learning with Small Data

Lorenzo Brigato, Luca Iocchi

Auto-TLDR; Low-Complex Neural Networks for Small Data Conditions

Abstract Slides Poster Similar

In this work, we perform a wide variety of experiments with different Deep Learning architectures in small data conditions. We show that model complexity is a critical factor when only a few samples per class are available. Differently from the literature, we improve the state of the art using low complexity models. We show that standard convolutional neural networks with relatively few parameters are effective in this scenario. In many of our experiments, low complexity models outperform state-of-the-art architectures. Moreover, we propose a novel network that uses an unsupervised loss to regularize its training. Such architecture either improves the results either performs comparably well to low capacity networks. Surprisingly, experiments show that the dynamic data augmentation pipeline is not beneficial in this particular domain. Statically augmenting the dataset might be a promising research direction while dropout maintains its role as a good regularizer.

Verifying the Causes of Adversarial Examples

Honglin Li, Yifei Fan, Frieder Ganz, Tony Yezzi, Payam Barnaghi

Auto-TLDR; Exploring the Causes of Adversarial Examples in Neural Networks

Abstract Slides Poster Similar

The robustness of neural networks is challenged by adversarial examples that contain almost imperceptible perturbations to inputs which mislead a classifier to incorrect outputs in high confidence. Limited by the extreme difficulty in examining a high-dimensional image space thoroughly, research on explaining and justifying the causes of adversarial examples falls behind studies on attacks and defenses. In this paper, we present a collection of potential causes of adversarial examples and verify (or partially verify) them through carefully-designed controlled experiments. The major causes of adversarial examples include model linearity, one-sum constraint, and geometry of the categories. To control the effect of those causes, multiple techniques are applied such as $L_2$ normalization, replacement of loss functions, construction of reference datasets, and novel models using multi-layer perceptron probabilistic neural networks (MLP-PNN) and density estimation (DE). Our experiment results show that geometric factors tend to be more direct causes and statistical factors magnify the phenomenon, especially for assigning high prediction confidence. We hope this paper will inspire more studies to rigorously investigate the root causes of adversarial examples, which in turn provide useful guidance on designing more robust models.

Feature Extraction by Joint Robust Discriminant Analysis and Inter-Class Sparsity

Fadi Dornaika, Ahmad Khoder

Auto-TLDR; Robust Discriminant Analysis with Feature Selection and Inter-class Sparsity (RDA_FSIS)

Abstract Slides Similar

Feature extraction methods have been successfully applied to many real-world applications. The classical Linear Discriminant Analysis (LDA) and its variants are widely used as feature extraction methods. Although they have been used for different classification tasks, these methods have some shortcomings. The main one is that the projection axes obtained are not informative about the relevance of original features. In this paper, we propose a linear embedding method that merges two interesting properties: Robust LDA and inter-class sparsity. Furthermore, the targeted projection transformation focuses on the most discriminant original features. The proposed method is called Robust Discriminant Analysis with Feature Selection and Inter-class Sparsity (RDA_FSIS). Two kinds of sparsity are explicitly included in the proposed model. The first kind is obtained by imposing the $\ell_{2,1}$ constraint on the projection matrix in order to perform feature ranking. The second kind is obtained by imposing the inter-class sparsity constraint used for getting a common sparsity structure in each class. Comprehensive experiments on five real-world image datasets demonstrate the effectiveness and advantages of our framework over existing linear methods.

Kernel-based Graph Convolutional Networks

Hichem Sahbi

Auto-TLDR; Spatial Graph Convolutional Networks in Recurrent Kernel Hilbert Space

Abstract Slides Poster Similar

Learning graph convolutional networks (GCNs) is an emerging field which aims at generalizing deep learning to arbitrary non-regular domains. Most of the existing GCNs follow a neighborhood aggregation scheme, where the representation of a node is recursively obtained by aggregating its neighboring node representations using averaging or sorting operations. However, these operations are either ill-posed or weak to be discriminant or increase the number of training parameters and thereby the computational complexity and the risk of overfitting. In this paper, we introduce a novel GCN framework that achieves spatial graph convolution in a reproducing kernel Hilbert space. The latter makes it possible to design, via implicit kernel representations, convolutional graph filters in a high dimensional and more discriminating space without increasing the number of training parameters. The particularity of our GCN model also resides in its ability to achieve convolutions without explicitly realigning nodes in the receptive fields of the learned graph filters with those of the input graphs, thereby making convolutions permutation agnostic and well defined. Experiments conducted on the challenging task of skeleton-based action recognition show the superiority of the proposed method against different baselines as well as the related work.

Tensorized Feature Spaces for Feature Explosion

Ravdeep Pasricha, Pravallika Devineni, Evangelos Papalexakis, Ramakrishnan Kannan

Auto-TLDR; Tensor Rank Decomposition for Hyperspectral Image Classification

Abstract Slides Poster Similar

In this paper, we present a novel framework that uses tensor factorization to generate richer feature spaces for pixel classification in hyperspectral images. In particular, we assess the performance of different tensor rank decomposition methods as compared to the traditional kernel-based approaches for the hyperspectral image classification problem. We propose ORION, which takes as input a hyperspectral image tensor and a rank and outputs an enhanced feature space from the factor matrices of the decomposed tensor. Our method is a feature explosion technique that inherently maps low dimensional input space in R^K to high dimensional space in R^R, where R >> K, say in the order of 1000x, like a kernel. We show how the proposed method exploits the multi-linear structure of hyperspectral three-dimensional tensor. We demonstrate the effectiveness of our method with experiments on three publicly available hyperspectral datasets with labeled pixels and compare their classification performance against traditional linear and non-linear supervised learning methods such as SVM with Linear, Polynomial, RBF kernels, and the Multi-Layer Perceptron model. Finally, we explore the relationship between the rank of the tensor decomposition and the classification accuracy using several hyperspectral datasets with ground truth.

Hierarchical Routing Mixture of Experts

Wenbo Zhao, Yang Gao, Shahan Ali Memon, Bhiksha Raj, Rita Singh

Auto-TLDR; A Binary Tree-structured Hierarchical Routing Mixture of Experts for Regression

Abstract Slides Poster Similar

In regression tasks the distribution of the data is often too complex to be fitted by a single model. In contrast, partition-based models are developed where data is divided and fitted by local models. These models partition the input space and do not leverage the input-output dependency of multimodal-distributed data, and strong local models are needed to make good predictions. Addressing these problems, we propose a binary tree-structured hierarchical routing mixture of experts (HRME) model that has classifiers as non-leaf node experts and simple regression models as leaf node experts. The classifier nodes jointly soft-partition the input-output space based on the natural separateness of multimodal data. This enables simple leaf experts to be effective for prediction. Further, we develop a probabilistic framework for the HRME model, and propose a recursive Expectation-Maximization (EM) based algorithm to learn both the tree structure and the expert models. Experiments on a collection of regression tasks validate the effectiveness of our method compared to a variety of other regression models.

Emerging Relation Network and Task Embedding for Multi-Task Regression Problems

Schreiber Jens, Bernhard Sick

Auto-TLDR; A Comparative Study of Multi-Task Learning for Non-linear Time Series Problems

Abstract Slides Poster Similar

Multi-Task learning (MTL) provides state-of-the-art results in many applications of computer vision and natural language processing. In contrast to single-task learning (STL), MTL allows for leveraging knowledge between related tasks improving prediction results on all tasks. However, there is a limited number of comparative studies applied to MTL architectures for regression and time series problems taking recent advances of MTL into account. An intriguing, non-linear time-series problem are day ahead forecasts of the expected power generation for renewable power plants. Therefore, the main contribution of this article is a comparative study of the following recent and relevant MTL architectures: Hard-parameter sharing, cross-stitch network, and sluice network (SN). They are compared to a multi-layer peceptron (MLP) model of similar size in an STL setting. As a additional contribution, we provide a simple, yet practical approach to model task specific information through an embedding layer in an MLP, referred to as task embedding. Further, we contribute a new MTL architecture named emerging relation network (ERN), which can be considered as an extension of the SN. For a solar power dataset, the task embedding achieves the best mean improvement with 8.2%. For two wind and one additional solar dataset, the ERN is the best MTL architecture with improvements up to 11.3%.

Rethinking of Deep Models Parameters with Respect to Data Distribution

Shitala Prasad, Dongyun Lin, Yiqun Li, Sheng Dong, Zaw Min Oo

Auto-TLDR; A progressive stepwise training strategy for deep neural networks

Abstract Slides Poster Similar

The performance of deep learning models are driven by various parameters but to tune all of them every time, for every dataset, is a heuristic practice. In this paper, unlike the common practice of decaying the learning rate, we propose a step-wise training strategy where the learning rate and the batch size are tuned based on the dataset size. Here, the given dataset size is progressively increased during the training to boost the network performance without saturating the learning curve, after certain epochs. We conducted extensive experiments on multiple networks and datasets to validate the proposed training strategy. The experimental results proves our hypothesis that the learning rate, the batch size and the data size are interrelated and can improve the network accuracy if an optimal progressive stepwise training strategy is applied. The proposed strategy also the overall training computational cost is reduced.

Resource-efficient DNNs for Keyword Spotting using Neural Architecture Search and Quantization

David Peter, Wolfgang Roth, Franz Pernkopf

Auto-TLDR; Neural Architecture Search for Keyword Spotting in Limited Resource Environments

Abstract Slides Poster Similar

This paper introduces neural architecture search (NAS) for the automatic discovery of small models for keyword spotting (KWS) in limited resource environments. We employ a differentiable NAS approach to optimize the structure of convolutional neural networks (CNNs) to meet certain memory constraints for storing weights as well as constraints on the number of operations per inference. Using NAS only, we were able to obtain a highly efficient model with 95.6% accuracy on the Google speech commands dataset with 494.8 kB of memory usage and 19.6 million operations. Additionally, weight quantization is used to reduce the memory consumption even further. We show that weight quantization to low bit-widths (e.g. 1 bit) can be used without substantial loss in accuracy. By increasing the number of input features from 10 MFCC to 20 MFCC we were able to increase the accuracy to 96.6% at 340.1 kB of memory usage and 27.1 million operations.

Can Data Placement Be Effective for Neural Networks Classification Tasks? Introducing the Orthogonal Loss

Brais Cancela, Veronica Bolon-Canedo, Amparo Alonso-Betanzos

Auto-TLDR; Spatial Placement for Neural Network Training Loss Functions

Abstract Slides Poster Similar

Traditionally, a Neural Network classification training loss function follows the same principle: minimizing the distance between samples that belong to the same class, while maximizing the distance to the other classes. There are no restrictions on the spatial placement of deep features (last layer input). This paper addresses this issue when dealing with Neural Networks, providing a set of loss functions that are able to train a classifier by forcing the deep features to be projected over a predefined orthogonal basis. Experimental results shows that these `data placement' functions can overcome the training accuracy provided by the classic cross-entropy loss function.

Progressive Learning Algorithm for Efficient Person Re-Identification

Zhen Li, Hanyang Shao, Liang Niu, Nian Xue

Auto-TLDR; Progressive Learning Algorithm for Large-Scale Person Re-Identification

Abstract Slides Poster Similar

This paper studies the problem of Person Re-Identification (ReID) for large-scale applications. Recent research efforts have been devoted to building complicated part models, which introduce considerably high computational cost and memory consumption, inhibiting its practicability in large-scale applications. This paper aims to develop a novel learning strategy to find efficient feature embeddings while maintaining the balance of accuracy and model complexity. More specifically, we find by enhancing the classical triplet loss together with cross-entropy loss, our method can explore the hard examples and build a discriminant feature embedding yet compact enough for large-scale applications. Our method is carried out progressively using Bayesian optimization, and we call it the Progressive Learning Algorithm (PLA). Extensive experiments on three large-scale datasets show that our PLA is comparable or better than the state-of-the-arts. Especially, on the challenging Market-1501 dataset, we achieve Rank-1=94.7\%/mAP=89.4\% while saving at least 30\% parameters than strong part models.

Quantifying Model Uncertainty in Inverse Problems Via Bayesian Deep Gradient Descent

Riccardo Barbano, Chen Zhang, Simon Arridge, Bangti Jin

Auto-TLDR; Bayesian Neural Networks for Inverse Reconstruction via Bayesian Knowledge-Aided Computation

Abstract Slides Poster Similar

Recent advances in reconstruction methods for inverse problems leverage powerful data-driven models, e.g., deep neural networks. These techniques have demonstrated state-of-the-art performances for several imaging tasks, but they often do not provide uncertainty on the obtained reconstructions. In this work, we develop a novel scalable data-driven knowledge-aided computational framework to quantify the model uncertainty via Bayesian neural networks. The approach builds on and extends deep gradient descent, a recently developed greedy iterative training scheme, and recasts it within a probabilistic framework. Scalability is achieved by being hybrid in the architecture: only the last layer of each block is Bayesian, while the others remain deterministic, and by being greedy in training. The framework is showcased on one representative medical imaging modality, viz. computed tomography with either sparse view or limited view data, and exhibits competitive performance with respect to state-of-the-art benchmarks, e.g., total variation, deep gradient descent and learned primal-dual.

Local Clustering with Mean Teacher for Semi-Supervised Learning

Zexi Chen, Benjamin Dutton, Bharathkumar Ramachandra, Tianfu Wu, Ranga Raju Vatsavai

Auto-TLDR; Local Clustering for Semi-supervised Learning

Abstract Slides Similar

The Mean Teacher (MT) model of Tarvainen and Valpola has shown favorable performance on several semi-supervised benchmark datasets. MT maintains a teacher model's weights as the exponential moving average of a student model's weights and minimizes the divergence between their probability predictions under diverse perturbations of the inputs. However, MT is known to suffer from confirmation bias, that is, reinforcing incorrect teacher model predictions. In this work, we propose a simple yet effective method called Local Clustering (LC) to mitigate the effect of confirmation bias. In MT, each data point is considered independent of other points during training; however, data points are likely to be close to each other in feature space if they share similar features. Motivated by this, we cluster data points locally by minimizing the pairwise distance between neighboring data points in feature space. Combined with a standard classification cross-entropy objective on labeled data points, the misclassified unlabeled data points are pulled towards high-density regions of their correct class with the help of their neighbors, thus improving model performance. We demonstrate on semi-supervised benchmark datasets SVHN and CIFAR-10 that adding our LC loss to MT yields significant improvements compared to MT and performance comparable to the state of the art in semi-supervised learning.

Meta Soft Label Generation for Noisy Labels

Görkem Algan, Ilkay Ulusoy

Auto-TLDR; MSLG: Meta-Learning for Noisy Label Generation

Abstract Slides Poster Similar

The existence of noisy labels in the dataset causes significant performance degradation for deep neural networks (DNNs). To address this problem, we propose a Meta Soft Label Generation algorithm called MSLG, which can jointly generate soft labels using meta-learning techniques and learn DNN parameters in an end-to-end fashion. Our approach adapts the meta-learning paradigm to estimate optimal label distribution by checking gradient directions on both noisy training data and noise-free meta-data. In order to iteratively update soft labels, meta-gradient descent step is performed on estimated labels, which would minimize the loss of noise-free meta samples. In each iteration, the base classifier is trained on estimated meta labels. MSLG is model-agnostic and can be added on top of any existing model at hand with ease. We performed extensive experiments on CIFAR10, Clothing1M and Food101N datasets. Results show that our approach outperforms other state-of-the-art methods by a large margin. Our code is available at \url{https://github.com/gorkemalgan/MSLG_noisy_label}.

Trainable Spectrally Initializable Matrix Transformations in Convolutional Neural Networks

Michele Alberti, Angela Botros, Schuetz Narayan, Rolf Ingold, Marcus Liwicki, Mathias Seuret

Auto-TLDR; Trainable and Spectrally Initializable Matrix Transformations for Neural Networks

Abstract Slides Poster Similar

In this work, we introduce a new architectural component to Neural Networks (NN), i.e., trainable and spectrally initializable matrix transformations on feature maps. While previous literature has already demonstrated the possibility of adding static spectral transformations as feature processors, our focus is on more general trainable transforms. We study the transforms in various architectural configurations on four datasets of different nature: from medical (ColorectalHist, HAM10000) and natural (Flowers) images to historical documents (CB55). With rigorous experiments that control for the number of parameters and randomness, we show that networks utilizing the introduced matrix transformations outperform vanilla neural networks. The observed accuracy increases appreciably across all datasets. In addition, we show that the benefit of spectral initialization leads to significantly faster convergence, as opposed to randomly initialized matrix transformations. The transformations are implemented as auto-differentiable PyTorch modules that can be incorporated into any neural network architecture. The entire code base is open-source.

Adversarial Encoder-Multi-Task-Decoder for Multi-Stage Processes

Andre Mendes, Julian Togelius, Leandro Dos Santos Coelho

Auto-TLDR; Multi-Task Learning and Semi-Supervised Learning for Multi-Stage Processes

Abstract Similar

In multi-stage processes, decisions occur in an ordered sequence of stages. Early stages usually have more observations with general information (easier/cheaper to collect), while later stages have fewer observations but more specific data. This situation can be represented by a dual funnel structure, in which the sample size decreases from one stage to the other while the information increases. Training classifiers in this scenario is challenging since information in the early stages may not contain distinct patterns to learn (underfitting). In contrast, the small sample size in later stages can cause overfitting. We address both cases by introducing a framework that combines adversarial autoencoders (AAE), multi-task learning (MTL), and multi-label semi-supervised learning (MLSSL). We improve the decoder of the AAE with an MTL component so it can jointly reconstruct the original input and use feature nets to predict the features for the next stages. We also introduce a sequence constraint in the output of an MLSSL classifier to guarantee the sequential pattern in the predictions. Using real-world data from different domains (selection process, medical diagnosis), we show that our approach outperforms other state-of-the-art methods.

Exploiting Non-Linear Redundancy for Neural Model Compression

Muhammad Ahmed Shah, Raphael Olivier, Bhiksha Raj

Auto-TLDR; Compressing Deep Neural Networks with Linear Dependency

Abstract Slides Poster Similar

Deploying deep learning models with millions, even billions, of parameters is challenging given real world memory, power and compute constraints. In an effort to make these models more practical, in this paper, we propose a novel model compression approach that exploits linear dependence between the activations in a layer to eliminate entire structural units (neurons/convolutional filters). Our approach also adjusts the weights of the layer in a manner that is provably lossless while training if the removed neuron was perfectly predictable. We combine this approach with an annealing algorithm that may be applied during training, or even on a trained model, and demonstrate, using popular datasets, that our technique can reduce the parameters of VGG and AlexNet by more than 97\% on \cifar, 85\% on \caltech, and 19\% on ImageNet at less than 2\% loss in accuracy. Furthermore, we provide theoretical results showing that in overparametrized, locally linear (ReLU) neural networks where redundant features exist, and with correct hyperparameter selection, our method is indeed able to capture and suppress those dependencies.

More Correlations Better Performance: Fully Associative Networks for Multi-Label Image Classification

Yaning Li, Liu Yang

Auto-TLDR; Fully Associative Network for Fully Exploiting Correlation Information in Multi-Label Classification

Abstract Slides Poster Similar

Recent researches demonstrate that correlation modeling plays a key role in high-performance multi-label classification methods. However, existing methods do not take full advantage of correlation information, especially correlations in feature and label spaces of each image, which limits the performance of correlation-based multi-label classification methods. With more correlations considered, in this study, a Fully Associative Network (FAN) is proposed for fully exploiting correlation information, which involves both visual feature and label correlations. Specifically, FAN introduces a robust covariance pooling to summarize convolution features as global image representation for capturing feature correlation in the multi-label task. Moreover, it constructs an effective label correlation matrix based on a re-weighted scheme, which is fed into a graph convolution network for capturing label correlation. Then, correlation between covariance representations (i.e., feature correlation ) and the outputs of GCN (i.e., label correlation) are modeled for final prediction. Experimental results on two datasets illustrate the effectiveness and efficiency of our proposed FAN compared with state-of-the-art methods.

On the Information of Feature Maps and Pruning of Deep Neural Networks

Mohammadreza Soltani, Suya Wu, Jie Ding, Robert Ravier, Vahid Tarokh

Auto-TLDR; Compressing Deep Neural Models Using Mutual Information

Abstract Slides Poster Similar

A technique for compressing deep neural models achieving competitive performance to state-of-the-art methods is proposed. The approach utilizes the mutual information between the feature maps and the output of the model in order to prune the redundant layers of the network. Extensive numerical experiments on both CIFAR-10, CIFAR-100, and Tiny ImageNet data sets demonstrate that the proposed method can be effective in compressing deep models, both in terms of the numbers of parameters and operations. For instance, by applying the proposed approach to DenseNet model with 0.77 million parameters and 293 million operations for classification of CIFAR-10 data set, a reduction of 62.66% and 41.00% in the number of parameters and the number of operations are respectively achieved, while increasing the test error only by less than 1%.

Nonlinear Ranking Loss on Riemannian Potato Embedding

Byung Hyung Kim, Yoonje Suh, Honggu Lee, Sungho Jo

Auto-TLDR; Riemannian Potato for Rank-based Metric Learning

Abstract Slides Poster Similar

We propose a rank-based metric learning method by leveraging a concept of the Riemannian Potato for better separating non-linear data. By exploring the geometric properties of Riemannian manifolds, the proposed loss function optimizes the measure of dispersion using the distribution of Riemannian distances between a reference sample and neighbors and builds a ranked list according to the similarities. We show the proposed function can learn a hypersphere for each class, preserving the similarity structure inside it on Riemannian manifold. As a result, compared with Euclidean distance-based metric, our method can further jointly reduce the intra-class distances and enlarge the inter-class distances for learned features, consistently outperforming state-of-the-art methods on three widely used non-linear datasets.

Contextual Classification Using Self-Supervised Auxiliary Models for Deep Neural Networks

Sebastian Palacio, Philipp Engler, Jörn Hees, Andreas Dengel

Auto-TLDR; Self-Supervised Autogenous Learning for Deep Neural Networks

Abstract Slides Poster Similar

Classification problems solved with deep neural networks (DNNs) typically rely on a closed world paradigm, and optimize over a single objective (e.g., minimization of the cross- entropy loss). This setup dismisses all kinds of supporting signals that can be used to reinforce the existence or absence of particular patterns. The increasing need for models that are interpretable by design makes the inclusion of said contextual signals a crucial necessity. To this end, we introduce the notion of Self-Supervised Autogenous Learning (SSAL). A SSAL objective is realized through one or more additional targets that are derived from the original supervised classification task, following architectural principles found in multi-task learning. SSAL branches impose low-level priors into the optimization process (e.g., grouping). The ability of using SSAL branches during inference, allow models to converge faster, focusing on a richer set of class-relevant features. We equip state-of-the-art DNNs with SSAL objectives and report consistent improvements for all of them on CIFAR100 and Imagenet. We show that SSAL models outperform similar state-of-the-art methods focused on contextual loss functions, auxiliary branches and hierarchical priors.

Class-Incremental Learning with Pre-Allocated Fixed Classifiers

Federico Pernici, Matteo Bruni, Claudio Baecchi, Francesco Turchini, Alberto Del Bimbo

Auto-TLDR; Class-Incremental Learning with Pre-allocated Output Nodes for Fixed Classifier

Abstract Slides Poster Similar

In class-incremental learning, a learning agent faces a stream of data with the goal of learning new classes while not forgetting previous ones. Neural networks are known to suffer under this setting, as they forget previously acquired knowledge. To address this problem, effective methods exploit past data stored in an episodic memory while expanding the final classifier nodes to accommodate the new classes. In this work, we substitute the expanding classifier with a novel fixed classifier in which a number of pre-allocated output nodes are subject to the classification loss right from the beginning of the learning phase. Contrarily to the standard expanding classifier, this allows: (a) the output nodes of future unseen classes to firstly see negative samples since the beginning of learning together with the positive samples that incrementally arrive; (b) to learn features that do not change their geometric configuration as novel classes are incorporated in the learning model. Experiments with public datasets show that the proposed approach is as effective as the expanding classifier while exhibiting intriguing properties of internal feature representation that are otherwise not-existent. Our ablation study on pre-allocating a large number of classes further validates the approach.

Probability Guided Maxout

Claudio Ferrari, Stefano Berretti, Alberto Del Bimbo

Auto-TLDR; Probability Guided Maxout for CNN Training

Abstract Slides Poster Similar

In this paper, we propose an original CNN training strategy that brings together ideas from both dropout-like regularization methods and solutions that learn discriminative features. We propose a dropping criterion that, differently from dropout and its variants, is deterministic rather than random. It grounds on the empirical evidence that feature descriptors with larger $L2$-norm and highly-active nodes are strongly correlated to confident class predictions. Thus, our criterion guides towards dropping a percentage of the most active nodes of the descriptors, proportionally to the estimated class probability. We simultaneously train a per-sample scaling factor to balance the expected output across training and inference. This further allows us to keep high the descriptor's L2-norm, which we show enforces confident predictions. The combination of these two strategies resulted in our ``Probability Guided Maxout'' solution that acts as a training regularizer. We prove the above behaviors by reporting extensive image classification results on the CIFAR10, CIFAR100, and Caltech256 datasets.

Interpretable Structured Learning with Sparse Gated Sequence Encoder for Protein-Protein Interaction Prediction

Kishan K C, Feng Cui, Anne Haake, Rui Li

Auto-TLDR; Predicting Protein-Protein Interactions Using Sequence Representations

Abstract Slides Poster Similar

Predicting protein-protein interactions (PPIs) by learning informative representations from amino acid sequences is a challenging yet important problem in biology. Although various deep learning models in Siamese architecture have been proposed to model PPIs from sequences, these methods are computationally expensive for a large number of PPIs due to the pairwise encoding process. Furthermore, these methods are difficult to interpret because of non-intuitive mappings from protein sequences to their sequence representation. To address these challenges, we present a novel deep framework to model and predict PPIs from sequence alone. Our model incorporates a bidirectional gated recurrent unit to learn sequence representations by leveraging contextualized and sequential information from sequences. We further employ a sparse regularization to model long-range dependencies between amino acids and to select important amino acids (protein motifs), thus enhancing interpretability. Besides, the novel design of the encoding process makes our model computationally efficient and scalable to an increasing number of interactions. Experimental results on up-to-date interaction datasets demonstrate that our model achieves superior performance compared to other state-of-the-art methods. Literature-based case studies illustrate the ability of our model to provide biological insights to interpret the predictions.

Budgeted Batch Mode Active Learning with Generalized Cost and Utility Functions

Arvind Agarwal, Shashank Mujumdar, Nitin Gupta, Sameep Mehta

Auto-TLDR; Active Learning Based on Utility and Cost Functions

Abstract Slides Poster Similar

Active learning reduces the labeling cost by actively querying labels for the most valuable data points. Typical active learning methods select the most informative examples one-at-a-time, their batch variants exist where a set of most informative points are selected. These points are selected in such a way that when added to the training data along with their labels, they provide maximum benefit to the underlying model. In this paper, we present a learning framework that actively selects optimal set of examples (in a batch) within a given budget, based on given utility and cost functions. The framework is generic enough to incorporate any utility and any cost function defined on a set of examples. Furthermore, we propose a novel utility function based on the Facility Location problem that considers three important characteristics of utility i.e., diversity, density and point utility. We also propose a novel cost function, by formulating the cost computation problem as an optimization problem, the solution to which turns out to be the minimum spanning tree. Thus, our framework provides the optimal batch of points within the given budget based on the cost and utility functions. We evaluate our method on several data sets and show its superior performance over baseline methods.