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.

The Effect of Multi-Step Methods on Overestimation in Deep Reinforcement Learning

Lingheng Meng, Rob Gorbet, Dana Kulić

Auto-TLDR; Multi-Step DDPG for Deep Reinforcement Learning

Abstract Slides Poster

Multi-step (also called n-step) methods in reinforcement learning (RL) have been shown to be more efficient than the 1-step method due to faster propagation of the reward signal, both theoretically and empirically, in tasks exploiting tabular representation of the value-function. Recently, research in Deep Reinforcement Learning (DRL) also shows that multi-step methods improve learning speed and final performance in applications where the value-function and policy are represented with deep neural networks. However, there is a lack of understanding about what is actually contributing to the boost of performance. In this work, we analyze the effect of multi-step methods on alleviating the overestimation problem in DRL, where multi-step experiences are sampled from a replay buffer. Specifically building on top of Deep Deterministic Policy Gradient (DDPG), we experiment with Multi-step DDPG (MDDPG), where different step sizes are manually set, and with a variant called Mixed Multi-step DDPG (MMDDPG) where an average over different multi-step backups is used as target Q-value. Empirically, we show that both MDDPG and MMDDPG are significantly less affected by the overestimation problem than DDPG with 1-step backup, which consequently results in better final performance and learning speed. We also discuss the advantages and disadvantages of different ways to do multi-step expansion in order to reduce approximation error, and expose the tradeoff between overestimation and underestimation that underlies offline multi-step methods. Finally, we compare the computational resource needs of TD3 and our proposed methods, since they show comparable final performance and learning speed.

Similar papers

A Bayesian Approach to Reinforcement Learning of Vision-Based Vehicular Control

Zahra Gharaee, Karl Holmquist, Linbo He, Michael Felsberg

Auto-TLDR; Bayesian Reinforcement Learning for Autonomous Driving

Abstract Slides Poster Similar

In this paper, we present a state-of-the-art reinforcement learning method for autonomous driving. Our approach employs temporal difference learning in a Bayesian framework to learn vehicle control signals from sensor data. The agent has access to images from a forward facing camera, which are pre-processed to generate semantic segmentation maps. We trained our system using both ground truth and estimated semantic segmentation input. Based on our observations from a large set of experiments, we conclude that training the system on ground truth input data leads to better performance than training the system on estimated input even if estimated input is used for evaluation. The system is trained and evaluated in a realistic simulated urban environment using the CARLA simulator. The simulator also contains a benchmark that allows for comparing to other systems and methods. The required training time of the system is shown to be lower and the performance on the benchmark superior to competing approaches.

Learning from Learners: Adapting Reinforcement Learning Agents to Be Competitive in a Card Game

Pablo Vinicius Alves De Barros, Ana Tanevska, Alessandra Sciutti

Auto-TLDR; Adaptive Reinforcement Learning for Competitive Card Games

Abstract Slides Poster Similar

Learning how to adapt to complex and dynamic environments is one of the most important factors that contribute to our intelligence. Endowing artificial agents with this ability is not a simple task, particularly in competitive scenarios. In this paper, we present a broad study on how popular reinforcement learning algorithms can be adapted and implemented to learn and to play a real-world implementation of a competitive multiplayer card game. We propose specific training and validation routines for the learning agents, in order to evaluate how the agents learn to be competitive and explain how they adapt to each others' playing style. Finally, we pinpoint how the behavior of each agent derives from their learning style and create a baseline for future research on this scenario.

AVD-Net: Attention Value Decomposition Network for Deep Multi-Agent Reinforcement Learning

Zhang Yuanxin, Huimin Ma, Yu Wang

Auto-TLDR; Attention Value Decomposition Network for Cooperative Multi-agent Reinforcement Learning

Abstract Slides Poster Similar

Multi-agent reinforcement learning (MARL) is of importance for variable real-world applications but remains more challenges like stationarity and scalability. While recently value function factorization methods have obtained empirical good results in cooperative multi-agent environment, these works mostly focus on the decomposable learning structures. Inspired by the application of attention mechanism in machine translation and other related domains, we propose an attention based approach called attention value decomposition network (AVD-Net), which capitalizes on the coordination relations between agents. AVD-Net employs centralized training with decentralized execution (CTDE) paradigm, which factorizes the joint action-value functions with only local observations and actions of agents. Our method is evaluated on multi-agent particle environment (MPE) and StarCraft micromanagement environment (SMAC). The experiment results show the strength of our approach compared to existing methods with state-of-the-art performance in cooperative scenarios.

Deep Reinforcement Learning on a Budget: 3D Control and Reasoning without a Supercomputer

Edward Beeching, Jilles Steeve Dibangoye, Olivier Simonin, Christian Wolf

Auto-TLDR; Deep Reinforcement Learning in Mobile Robots Using 3D Environment Scenarios

Abstract Slides Poster Similar

An important goal of research in Deep Reinforcement Learning in mobile robotics is to train agents capableof solving complex tasks, which require a high level of scene understanding and reasoning from an egocentric perspective.When trained from simulations, optimal environments should satisfy a currently unobtainable combination of high-fidelity photographic observations, massive amounts of different environment configurations and fast simulation speeds. In this paper we argue that research on training agents capable of complex reasoning can be simplified by decoupling from the requirement of high fidelity photographic observations. We present a suite of tasks requiring complex reasoning and exploration in continuous,partially observable 3D environments. The objective is to provide challenging scenarios and a robust baseline agent architecture that can be trained on mid-range consumer hardware in under 24h. Our scenarios combine two key advantages: (i) they are based on a simple but highly efficient 3D environment (ViZDoom)which allows high speed simulation (12000fps); (ii) the scenarios provide the user with a range of difficulty settings, in order to identify the limitations of current state of the art algorithms and network architectures. We aim to increase accessibility to the field of Deep-RL by providing baselines for challenging scenarios where new ideas can be iterated on quickly. We argue that the community should be able to address challenging problems in reasoning of mobile agents without the need for a large compute infrastructure.

Low Dimensional State Representation Learning with Reward-Shaped Priors

Nicolò Botteghi, Ruben Obbink, Daan Geijs, Mannes Poel, Beril Sirmacek, Christoph Brune, Abeje Mersha, Stefano Stramigioli

Auto-TLDR; Unsupervised Learning for Unsupervised Reinforcement Learning in Robotics

Abstract Poster Similar

Reinforcement Learning has been able to solve many complicated robotics tasks without any need of feature engineering in an end-to-end fashion. However, learning the optimal policy directly from the sensory inputs, i.e the observations, often requires processing and storage of huge amount of data. In the context of robotics, the cost of data from real robotics hardware is usually very high, thus solutions that achieves high sample-efficiency are needed. We propose a method that aims at learning a mapping from the observations into a lower dimensional state space. This mapping is learned with unsupervised learning using loss functions shaped to incorporate prior knowledge of the environment and the task. Using the samples from the state space, the optimal policy is quickly and efficiently learned. We test the method on several mobile robot navigation tasks in simulation environment and also on a real robot.

Meta Learning Via Learned Loss

Sarah Bechtle, Artem Molchanov, Yevgen Chebotar, Edward Thomas Grefenstette, Ludovic Righetti, Gaurav Sukhatme, Franziska Meier

Auto-TLDR; meta-learning for learning parametric loss functions that generalize across different tasks and model architectures

Abstract Slides Similar

Typically, loss functions, regularization mechanisms and other important aspects of training parametric models are chosen heuristically from a limited set of options. In this paper, we take the first step towards automating this process, with the view of producing models which train faster and more robustly. Concretely, we present a meta-learning method for learning parametric loss functions that can generalize across different tasks and model architectures. We develop a pipeline for “meta-training” such loss functions, targeted at maximizing the performance of the model trained under them. The loss landscape produced by our learned losses significantly improves upon the original task-specific losses in both supervised and reinforcement learning tasks. Furthermore, we show that our meta-learning framework is flexible enough to incorporate additional information at meta-train time. This information shapes the learned loss function such that the environment does not need to provide this information during meta-test time.

Object-Oriented Map Exploration and Construction Based on Auxiliary Task Aided DRL

Junzhe Xu, Jianhua Zhang, Shengyong Chen, Honghai Liu

Auto-TLDR; Auxiliary Task Aided Deep Reinforcement Learning for Environment Exploration by Autonomous Robots

Abstract Similar

Environment exploration by autonomous robots through deep reinforcement learning (DRL) based methods has attracted more and more attention. However, existing methods usually focus on robot navigation to single or multiple fixed goals, while ignoring the perception and construction of external environments. In this paper, we propose a novel environment exploration task based on DRL, which requires a robot fast and completely perceives all objects of interest, and reconstructs their poses in a global environment map, as much as the robot can do. To this end, we design an auxiliary task aided DRL model, which is integrated with the auxiliary object detection and 6-DoF pose estimation components. The outcome of auxiliary tasks can improve the learning speed and robustness of DRL, as well as the accuracy of object pose estimation. Comprehensive experimental results on the indoor simulation platform AI2-THOR have shown the effectiveness and robustness of our method.

Can Reinforcement Learning Lead to Healthy Life?: Simulation Study Based on User Activity Logs

Masami Takahashi, Masahiro Kohjima, Takeshi Kurashima, Hiroyuki Toda

Auto-TLDR; Reinforcement Learning for Healthy Daily Life

Abstract Slides Poster Similar

The importance of developing an application based on intervention technology that leads to a healthier life is widely recognized. A challenging part of realizing the application is the need for planning, i.e., considering a user's health goal (e.g., sleep at 10:00 p.m. to get enough sleep), providing intervention at the appropriate timing to help the user achieve the goal. The reinforcement learning (RL) approach is well suited to this type of problem since it is a methodology for planning; RL finds the optimal strategy as that which maximizes future expected profit. The purpose of this study is to clarify the effects of intervention based on RL to support healthy daily life. Therefore, we (i) collect real daily activity data from participants, (ii) generate a user model that imitates the user's response to system interventions, (iii) examine valuable goals and design them as rewards in RL and (iv) obtain optimal intervention strategies by RL via simulations given a user model and goals. We evaluate a generated user model and verify by simulations whether our method could successfully achieve the goal. In addition, we analyze the cases that demonstrated higher probability of achieving the goal and report the features.

Detecting and Adapting to Crisis Pattern with Context Based Deep Reinforcement Learning

Eric Benhamou, David Saltiel Saltiel, Jean-Jacques Ohana Ohana, Jamal Atif Atif

Auto-TLDR; Deep Reinforcement Learning for Financial Crisis Detection and Dis-Investment

Abstract Slides Poster Similar

Deep reinforcement learning (DRL) has reached super human levels in complexes tasks like game solving (Go, StarCraft II), and autonomous driving. However, it remains an open question whether DRL can reach human level in applications to financial problems and in particular in detecting pattern crisis and consequently dis-investing. In this paper, we present an innovative DRL framework consisting in two sub-networks fed respectively with portfolio strategies past performances and standard deviation as well as additional contextual features. The second sub network plays an important role as it captures dependencies with common financial indicators features like risk aversion, economic surprise index and correlations between assets that allows taking into account context based information. We compare different network architectures either using layers of convolutions to reduce network's complexity or LSTM block to capture time dependency and whether previous allocations is important in the modeling. We also use adversarial training to make the final model more robust. Results on test set show this approach substantially over-performs traditional portfolio optimization methods like Markovitz and is able to detect and anticipate crisis like the current Covid one.

Self-Play or Group Practice: Learning to Play Alternating Markov Game in Multi-Agent System

Chin-Wing Leung, Shuyue Hu, Ho-Fung Leung

Auto-TLDR; Group Practice for Deep Reinforcement Learning

Abstract Slides Poster Similar

The research in reinforcement learning has achieved great success in strategic game playing. These successes are thanks to the incorporation of deep reinforcement learning (DRL) and Monte Carlo Tree Search (MCTS) to the agent trained under the self-play (SP) environment. By self-play, agents are provided with an incrementally more difficult curriculum which in turn facilitate learning. However, recent research suggests that agents trained via self-play may easily lead to getting stuck in local equilibria. In this paper, we consider a population of agents each independently learns to play an alternating Markov game (AMG). We propose a new training framework---group practice---for a population of decentralized RL agents. By group practice (GP), agents are assigned into multiple learning groups during training, for every episode of games, an agent is randomly paired up and practices with another agent in the learning group. The convergence result to the optimal value function and the Nash equilibrium are proved under the GP framework. Experimental study is conducted by applying GP to Q-learning algorithm and the deep Q-learning with Monte-Carlo tree search on the game of Connect Four and the game of Hex. We verify that GP is the more efficient training scheme than SP given the same amount of training. We also show that the learning effectiveness can even be improved when applying local grouping to agents.

Deep Reinforcement Learning for Autonomous Driving by Transferring Visual Features

Hongli Zhou, Guanwen Zhang, Wei Zhou

Auto-TLDR; Deep Reinforcement Learning for Autonomous Driving by Transferring Visual Features

Abstract Slides Poster Similar

Deep reinforcement learning (DRL) has achieved great success in processing vision-based driving tasks. However, the end-to-end training manner makes DRL agents suffer from overfitting training scenes. The agents easily fail to generalize to unseen environments. In this paper, we propose a deep reinforcement learning for autonomous driving by transferring visual features. We formulate the DRL training as a perception and control module and introduce adversarial training mechanism for autonomous driving. The perception module is able to extract invariant features between different domains through adversarial training. While the DRL agent can then be trained on the basis of low dimensional states. In this manner, the proposed approach enables trained agents to adapt to unseen environments by learning robust features invariant across various scenes. We evaluate the proposed approach by transferring visual features between different simulators. The experimental results demonstrate the driving policy trained in the source domain can be directly applied in the target domain, and achieve great efficient and effective performance for autonomous driving.

Trajectory Representation Learning for Multi-Task NMRDP Planning

Firas Jarboui, Vianney Perchet

Auto-TLDR; Exploring Non Markovian Reward Decision Processes for Reinforcement Learning

Abstract Slides Poster Similar

Expanding Non Markovian Reward Decision Processes (NMRDP) into Markov Decision Processes (MDP) enables the use of state of the art Reinforcement Learning (RL) techniques to identify optimal policies. In this paper an approach to exploring NMRDPs and expanding them into MDPs, without the prior knowledge of the reward structure, is proposed. The non Markovianity of the reward function is disentangled under the assumption that sets of similar and dissimilar trajectory batches can be sampled. More precisely, within the same batch, measuring the similarity between any couple of trajectories is permitted, although comparing trajectories from different batches is not possible. A modified version of the triplet loss is optimised to construct a representation of the trajectories under which rewards become Markovian.

AOAM: Automatic Optimization of Adjacency Matrix for Graph Convolutional Network

Yuhang Zhang, Hongshuai Ren, Jiexia Ye, Xitong Gao, Yang Wang, Kejiang Ye, Cheng-Zhong Xu

Auto-TLDR; Adjacency Matrix for Graph Convolutional Network in Non-Euclidean Space

Abstract Slides Poster Similar

Graph Convolutional Network (GCN) is adopted to tackle the problem of the convolution operation in non-Euclidean space. Although previous works on GCN have made some progress, one of their limitations is that their input Adjacency Matrix (AM) is designed manually and requires domain knowledge, which is cumbersome, tedious and error-prone. In addition, entries of this fixed Adjacency Matrix are generally designed as binary values (i.e., ones and zeros) which can not reflect more complex relationship between nodes. However, many applications require a weighted and dynamic Adjacency Matrix instead of an unweighted and fixed Adjacency Matrix. To this end, there are few works focusing on designing a more flexible Adjacency Matrix. In this paper, we propose an end-to-end algorithm to improve the GCN performance by focusing on the Adjacency Matrix. We first provide a calculation method that called node information entropy to update the matrix. Then, we analyze the search strategy in a continuous space and introduce the Deep Deterministic Policy Gradient (DDPG) method to overcome the demerit of the discrete space search. Finally, we integrate the GCN and reinforcement learning into an end-to-end framework. Our method can automatically define the adjacency matrix without artificial knowledge. At the same time, the proposed approach can deal with any size of the matrix and provide a better value for the network. Four popular datasets are selected to evaluate the capability of our algorithm. The method in this paper achieves the state-of-the-art performance on Cora and Pubmed datasets, respectively, with the accuracy of 84.6% and 81.6%.

Vacant Parking Space Detection Based on Task Consistency and Reinforcement Learning

Manh Hung Nguyen, Tzu-Yin Chao, Ching-Chun Huang

Auto-TLDR; Vacant Space Detection via Semantic Consistency Learning

Abstract Slides Poster Similar

In this paper, we proposed a novel task-consistency learning method that allows training a vacant space detection network (target task) based on the logistic consistency with the semantic outcomes from a naive flow-based motion behavior classifier (source task) in a parking lot. By well designing the reward mechanism upon semantic consistency, we show the possibility to train the target network in a reinforcement learning setting. Compared with conventional supervised detection methods, the major contribution of this work is to learn a vacant space detector via semantic consistency rather than supervised labels. The dynamic learning property may make the proposed detector been deployed in different lots easily without heavy training loads. The experiments show that based on the task consistency rewards from the motion behavior classifier, the vacant space detector can be trained successfully.

Rethinking Experience Replay: A Bag of Tricks for Continual Learning

Pietro Buzzega, Matteo Boschini, Angelo Porrello, Simone Calderara

Auto-TLDR; Experience Replay for Continual Learning: A Practical Approach

Abstract Slides Poster Similar

In Continual Learning, a Neural Network is trained on a stream of data whose distribution shifts over time. Under these assumptions, it is especially challenging to improve on classes appearing later in the stream while remaining accurate on previous ones. This is due to the infamous problem of catastrophic forgetting, which causes a quick performance degradation when the classifier focuses on learning new categories. Recent literature proposed various approaches to tackle this issue, often resorting to very sophisticated techniques. In this work, we show that naive rehearsal can be patched to achieve similar performance. We point out some shortcomings that restrain Experience Replay (ER) and propose five tricks to mitigate them. Experiments show that ER, thus enhanced, displays an accuracy gain of 51.2 and 26.9 percentage points on the CIFAR-10 and CIFAR-100 datasets respectively (memory buffer size 1000). As a result, it surpasses current state-of-the-art rehearsal-based methods.

A Novel Actor Dual-Critic Model for Remote Sensing Image Captioning

Ruchika Chavhan, Biplab Banerjee, Xiao Xiang Zhu, Subhasis Chaudhuri

Auto-TLDR; Actor Dual-Critic Training for Remote Sensing Image Captioning Using Deep Reinforcement Learning

Abstract Slides Poster Similar

We deal with the problem of generating textual captions from optical remote sensing (RS) images using the notion of deep reinforcement learning. Due to the high inter-class similarity in reference sentences describing remote sensing data, jointly encoding the sentences and images encourages prediction of captions that are semantically more precise than the ground truth in many cases. To this end, we introduce an Actor Dual-Critic training strategy where a second critic model is deployed in the form of an encoder-decoder RNN to encode the latent information corresponding to the original and generated captions. While all actor-critic methods use an actor to predict sentences for an image and a critic to provide rewards, our proposed encoder-decoder RNN guarantees high-level comprehension of images by sentence-to-image translation. We observe that the proposed model generates sentences on the test data highly similar to the ground truth and is successful in generating even better captions in many critical cases. Extensive experiments on the benchmark Remote Sensing Image Captioning Dataset (RSICD) and the UCM-captions dataset confirm the superiority of the proposed approach in comparison to the previous state-of-the-art where we obtain a gain of sharp increments in both the ROUGE-L and CIDEr measures.

Explore and Explain: Self-Supervised Navigation and Recounting

Roberto Bigazzi, Federico Landi, Marcella Cornia, Silvia Cascianelli, Lorenzo Baraldi, Rita Cucchiara

Auto-TLDR; Exploring a Photorealistic Environment for Explanation and Navigation

Abstract Slides Similar

Embodied AI has been recently gaining attention as it aims to foster the development of autonomous and intelligent agents. In this paper, we devise a novel embodied setting in which an agent needs to explore a previously unknown environment while recounting what it sees during the path. In this context, the agent needs to navigate the environment driven by an exploration goal, select proper moments for description, and output natural language descriptions of relevant objects and scenes. Our model integrates a novel self-supervised exploration module with penalty, and a fully-attentive captioning model for explanation. Also, we investigate different policies for selecting proper moments for explanation, driven by information coming from both the environment and the navigation. Experiments are conducted on photorealistic environments from the Matterport3D dataset and investigate the navigation and explanation capabilities of the agent as well as the role of their interactions.

ActionSpotter: Deep Reinforcement Learning Framework for Temporal Action Spotting in Videos

Guillaume Vaudaux-Ruth, Adrien Chan-Hon-Tong, Catherine Achard

Auto-TLDR; ActionSpotter: A Reinforcement Learning Algorithm for Action Spotting in Video

Abstract Slides Poster Similar

Action spotting has recently been proposed as an alternative to action detection and key frame extraction. However, the current state-of-the-art method of action spotting requires an expensive ground truth composed of the search sequences employed by human annotators spotting actions - a critical limitation. In this article, we propose to use a reinforcement learning algorithm to perform efficient action spotting using only the temporal segments from the action detection annotations, thus opening an interesting solution for video understanding. Experiments performed on THUMOS14 and ActivityNet datasets show that the proposed method, named ActionSpotter, leads to good results and outperforms state-of-the-art detection outputs redrawn for this application. In particular, the spotting mean Average Precision on THUMOS14 is significantly improved from 59.7% to 65.6% while skipping 23% of video.

Visual Object Tracking in Drone Images with Deep Reinforcement Learning

Derya Gözen, Sedat Ozer

Auto-TLDR; A Deep Reinforcement Learning based Single Object Tracker for Drone Applications

Abstract Slides Poster Similar

There is an increasing demand on utilizing camera equipped drones and their applications in many domains varying from agriculture to entertainment and from sports events to surveillance. In such drone applications, an essential and a common task is tracking an object of interest visually. Drone (or UAV) images have different properties when compared to the ground taken (natural) images and those differences introduce additional complexities to the existing object trackers to be directly applied on drone applications. Some important differences among those complexities include (i) smaller object sizes to be tracked and (ii) different orientations and viewing angles yielding different texture and features to be observed. Therefore, new algorithms trained on drone images are needed for the drone-based applications. In this paper, we introduce a deep reinforcement learning (RL) based single object tracker that tracks an object of interest in drone images by estimating a series of actions to find the location of the object in the next frame. This is the first work introducing a single object tracker using a deep RL-based technique for drone images. Our proposed solution introduces a novel reward function that aims to reduce the total number of actions taken to estimate the object's location in the next frame and also introduces a different backbone network to be used on low resolution images. Additionally, we introduce a set of new actions into the action library to better deal with the above-mentioned complexities. We compare our proposed solutions to a state of the art tracking algorithm from the recent literature and demonstrate up to 3.87\% improvement in precision and 3.6\% improvement in IoU values on the VisDrone2019 dataset. We also provide additional results on OTB-100 dataset and show up to 3.15\% improvement in precision on the OTB-100 dataset when compared to the same previous state of the art algorithm. Lastly, we analyze the ability to handle some of the challenges faced during tracking, including but not limited to occlusion, deformation, and scale variation for our proposed solutions.

Adaptive Remote Sensing Image Attribute Learning for Active Object Detection

Nuo Xu, Chunlei Huo, Chunhong Pan

Auto-TLDR; Adaptive Image Attribute Learning for Active Object Detection

Abstract Slides Similar

In recent years, deep learning methods bring incredible progress to the field of object detection. However, in the field of remote sensing image processing, existing methods neglect the relationship between imaging configuration and detection performance, and do not take into account the importance of detection performance feedback for improving image quality. Therefore, detection performance is limited by the passive nature of the conventional object detection framework. In order to solve the above limitations, this paper takes adaptive brightness adjustment and scale adjustment as examples, and proposes an active object detection method based on deep reinforcement learning. The goal of adaptive image attribute learning is to maximize the detection performance. With the help of active object detection and image attribute adjustment strategies, low-quality images can be converted into high-quality images, and the overall performance is improved without retraining the detector.

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.

On Embodied Visual Navigation in Real Environments through Habitat

Marco Rosano, Antonino Furnari, Luigi Gulino, Giovanni Maria Farinella

Auto-TLDR; Learning Navigation Policies on Real World Observations using Real World Images and Sensor and Actuation Noise

Abstract Slides Poster Similar

Visual navigation models based on deep learning can learn effective policies when trained on large amounts of visual observations through reinforcement learning. Unfortunately, collecting the required experience deploying a robotic platform in the real world is expensive and time-consuming. To deal with this limitation, several simulation platforms have been proposed in order to train visual navigation policies on virtual environments efficiently. Despite the advantages they offer, simulators present a limited realism in terms of appearance and physical dynamics, leading to navigation policies that do not generalize in the real world. In this paper, we propose a tool based on the Habitat simulator which exploits real world images of the environment, together with sensor and actuator noise models, to produce more realistic navigation episodes. We perform a range of experiments using virtual, real and images transformed with a simple domain adaptation approach. We also assess the impact of sensor and actuation noise on the navigation performance and investigate whether they allow to learn more robust navigation policies. We show that our tool can effectively help to train and evaluate navigation policies on real world observations without running navigation episodes in the real world.

Naturally Constrained Online Expectation Maximization

Daniela Pamplona, Antoine Manzanera

Auto-TLDR; Constrained Online Expectation-Maximization for Probabilistic Principal Components Analysis

Abstract Slides Poster Similar

With the rise of big data sets, learning algorithms must be adapted to piece-wise mechanisms in order to tackle time and memory costs of large scale calculations. Furthermore, for most learning embedded systems the input data are fed in a sequential and contingent manner: one by one, and possibly class by class. Thus, learning algorithms should not only run online but cope with time-varying, non-independent, and non-balanced training data for the system's entire life. Online Expectation-Maximization is a well-known algorithm for learning probabilistic models in real-time, due to its simplicity and convergence properties. However, these properties are only valid in the case of large, independent and identically distributed (iid) samples. In this paper, we propose to constraint the online Expectation-Maximization on the Fisher distance between the parameters. After the presentation of the algorithm, we make a thorough study of its use in Probabilistic Principal Components Analysis. First, we derive the update rules, then we analyse the effect of the constraint on major problems of online and sequential learning: convergence, forgetting and interference. Furthermore we use several algorithmic protocols: iid {\em vs} sequential data, and constraint parameters updated step-wise {\em vs} class-wise. Our results show that this constraint increases the convergence rate of online Expectation-Maximization, decreases forgetting and slightly introduces transfer learning.

Deep Next-Best-View Planner for Cross-Season Visual Route Classification

Kurauchi Kanya, Kanji Tanaka

Auto-TLDR; Active Visual Place Recognition using Deep Convolutional Neural Network

Abstract Slides Poster Similar

This paper addresses the problem of active visual place recognition (VPR) from a novel perspective of long-term autonomy. In our approach, a next-best-view (NBV) planner plans an optimal action-observation-sequence to maximize the expected cost-performance for a visual route classification task. A difficulty arises from the fact that the NBV planner is trained and tested in different domains (times of day, weather conditions, and seasons). Existing NBV methods may be confused and deteriorated by the domain-shifts, and require significant efforts for adapting them to a new domain. We address this issue by a novel deep convolutional neural network (DNN) -based NBV planner that does not require the adaptation. Our main contributions in this paper are summarized as follows: (1) We present a novel domain-invariant NBV planner that is specifically tailored for DNN-based VPR. (2) We formulate the active VPR as a POMDP problem and present a feasible solution to address the inherent intractability. Specifically, the probability distribution vector (PDV) output by the available DNN is used as a domain-invariant observation model without the need to retrain it. (3) We verify efficacy of the proposed approach through challenging cross-season VPR experiments, where it is confirmed that the proposed approach clearly outperforms the previous single-view-based or multi-view-based VPR in terms of VPR accuracy and/or action-observation-cost.

Improving Visual Question Answering Using Active Perception on Static Images

Theodoros Bozinis, Nikolaos Passalis, Anastasios Tefas

Auto-TLDR; Fine-Grained Visual Question Answering with Reinforcement Learning-based Active Perception

Abstract Slides Poster Similar

Visual Question Answering (VQA) is one of the most challenging emerging applications of deep learning. Providing powerful attention mechanisms is crucial for VQA, since the model must correctly identify the region of an image that is relevant to the question at hand. However, existing models analyze the input images at a fixed and typically small resolution, often leading to discarding valuable fine-grained details. To overcome this limitation, in this work we propose a reinforcement learning-based active perception approach that works by applying a series of transformation operations on the images (translation, zoom) in order to facilitate answering the question at hand. This allows for performing fine-grained analysis, effectively increasing the resolution at which the models process information. The proposed method is orthogonal to existing attention mechanisms and it can be combined with most existing VQA methods. The effectiveness of the proposed method is experimentally demonstrated on a challenging VQA dataset.

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.

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.

Dual-Memory Model for Incremental Learning: The Handwriting Recognition Use Case

Mélanie Piot, Bérangère Bourdoulous, Aurelia Deshayes, Lionel Prevost

Auto-TLDR; A dual memory model for handwriting recognition

Abstract Poster Similar

In this paper, we propose a dual memory model inspired by neural science. Short-term memory processes the data stream before integrating them into long-term memory, which generalizes. The use case is learning the ability to recognize handwriting. This begins with the learning of prototypical letters . It continues throughout life and gives the individual the ability to recognize increasingly varied handwriting. This second task is achieved by incrementally training our dual-memory model. We used a convolution network for encoding and random forests as the memory model. Indeed, the latter have the advantage of being easily enhanced to integrate new data and new classes. Performances on the MNIST database are very encouraging since they exceed 95\% and the complexity of the model remains reasonable.

Selecting Useful Knowledge from Previous Tasks for Future Learning in a Single Network

Feifei Shi, Peng Wang, Zhongchao Shi, Yong Rui

Auto-TLDR; Continual Learning with Gradient-based Threshold Threshold

Abstract Slides Poster Similar

Continual learning is able to learn new tasks incrementally while avoiding catastrophic forgetting. Recent work has shown that packing multiple tasks into a single network incrementally by iterative pruning and re-training network is a promising method. We build upon this idea and propose an improved version of PackNet, specifically, we propose a novel gradient-based threshold method to reuse the knowledge of the previous tasks selectively when learning new tasks. Our experiments on a variety of classification tasks and different network architectures demonstrate that our method obtains competitive results when compared to PackNet.

Improving Robotic Grasping on Monocular Images Via Multi-Task Learning and Positional Loss

William Prew, Toby Breckon, Magnus Bordewich, Ulrik Beierholm

Auto-TLDR; Improving grasping performance from monocularcolour images in an end-to-end CNN architecture with multi-task learning

Abstract Slides Poster Similar

In this paper we introduce two methods of improv-ing real-time objecting grasping performance from monocularcolour images in an end-to-end CNN architecture. The first isthe addition of an auxiliary task during model training (multi-task learning). Our multi-task CNN model improves graspingperformance from a baseline average of 72.04% to 78.14% onthe large Jacquard grasping dataset when performing a supple-mentary depth reconstruction task. The second is introducinga positional loss function that emphasises loss per pixel forsecondary parameters (gripper angle and width) only on points ofan object where a successful grasp can take place. This increasesperformance from a baseline average of 72.04% to 78.92% aswell as reducing the number of training epochs required. Thesemethods can be also performed in tandem resulting in a furtherperformance increase to 79.12%, while maintaining sufficientinference speed to enable processing at 50FPS

RLST: A Reinforcement Learning Approach to Scene Text Detection Refinement

Xuan Peng, Zheng Huang, Kai Chen, Jie Guo, Weidong Qiu

Auto-TLDR; Saccadic Eye Movements and Peripheral Vision for Scene Text Detection using Reinforcement Learning

Abstract Slides Poster Similar

Within the research of scene text detection, some previous work has already achieved significant accuracy and efficiency. However, most of the work was generally done without considering about the implicit relationship between detection and eye movements. In this paper, we propose a new method for scene text detection especially for its refinement based on reinforcement learning. The idea of this method is inspired by Saccadic Eye Movements and Peripheral Vision. A saccade makes it possible for humans to orient the gaze to the location where a visual object has appeared. Peripheral vision gathers visual information of surroundings which provides supplement to foveal vision during gazing. We propose a simple pipeline, imitating the way human eyes do a saccade and collect peripheral information, to locate scene text roughly and to refine multi-scale vision field iteratively using reinforcement learning. For both training and evaluation, we use ICDAR2015 Challenge 4 dataset as a base and design several criteria to measure the feasibility of our work.

P-DIFF: Learning Classifier with Noisy Labels Based on Probability Difference Distributions

Wei Hu, Qihao Zhao, Yangyu Huang, Fan Zhang

Auto-TLDR; P-DIFF: A Simple and Effective Training Paradigm for Deep Neural Network Classifier with Noisy Labels

Abstract Slides Poster Similar

Learning deep neural network (DNN) classifier with noisy labels is a challenging task because the DNN can easily over- fit on these noisy labels due to its high capability. In this paper, we present a very simple but effective training paradigm called P-DIFF, which can train DNN classifiers but obviously alleviate the adverse impact of noisy labels. Our proposed probability difference distribution implicitly reflects the probability of a training sample to be clean, then this probability is employed to re-weight the corresponding sample during the training process. P-DIFF can also achieve good performance even without prior- knowledge on the noise rate of training samples. Experiments on benchmark datasets also demonstrate that P-DIFF is superior to the state-of-the-art sample selection methods.

RSAC: Regularized Subspace Approximation Classifier for Lightweight Continuous Learning

Chih-Hsing Ho, Shang-Ho Tsai

Auto-TLDR; Regularized Subspace Approximation Classifier for Lightweight Continuous Learning

Abstract Slides Poster Similar

Continuous learning seeks to perform the learning on the data that arrives from time to time. While prior works have demonstrated several possible solutions, these approaches require excessive training time as well as memory usage. This is impractical for applications where time and storage are constrained, such as edge computing. In this work, a novel training algorithm, regularized subspace approximation classifier (RSAC), is proposed to achieve lightweight continuous learning. RSAC contains a feature reduction module and classifier module with regularization. Extensive experiments show that RSAC is more efficient than prior continuous learning works and outperforms these works on various experimental settings.

SAILenv: Learning in Virtual Visual Environments Made Simple

Enrico Meloni, Luca Pasqualini, Matteo Tiezzi, Marco Gori, Stefano Melacci

Auto-TLDR; SAILenv: A Simple and Customized Platform for Visual Recognition in Virtual 3D Environment

Abstract Slides Poster Similar

Recently, researchers in Machine Learning algorithms, Computer Vision scientists, engineers and others, showed a growing interest in 3D simulators as a mean to artificially create experimental settings that are very close to those in the real world. However, most of the existing platforms to interface algorithms with 3D environments are often designed to setup navigation-related experiments, to study physical interactions, or to handle ad-hoc cases that are not thought to be customized, sometimes lacking a strong photorealistic appearance and an easy-to-use software interface. In this paper, we present a novel platform, SAILenv, that is specifically designed to be simple and customizable, and that allows researchers to experiment visual recognition in virtual 3D scenes. A few lines of code are needed to interface every algorithm with the virtual world, and non-3D-graphics experts can easily customize the 3D environment itself, exploiting a collection of photorealistic objects. Our framework yields pixel-level semantic and instance labeling, depth, and, to the best of our knowledge, it is the only one that provides motion-related information directly inherited from the 3D engine. The client-server communication operates at a low level, avoiding the overhead of HTTP-based data exchanges. We perform experiments using a state-of-the-art object detector trained on real-world images, showing that it is able to recognize the photorealistic 3D objects of our environment. The computational burden of the optical flow compares favourably with the estimation performed using modern GPU-based convolutional networks or more classic implementations. We believe that the scientific community will benefit from the easiness and high-quality of our framework to evaluate newly proposed algorithms in their own customized realistic conditions.

P2 Net: Augmented Parallel-Pyramid Net for Attention Guided Pose Estimation

Luanxuan Hou, Jie Cao, Yuan Zhao, Haifeng Shen, Jian Tang, Ran He

Auto-TLDR; Parallel-Pyramid Net with Partial Attention for Human Pose Estimation

Abstract Slides Poster Similar

The target of human pose estimation is to determine the body parts and joint locations of persons in the image. Angular changes, motion blur and occlusion etc. in the natural scenes make this task challenging, while some joints are more difficult to be detected than others. In this paper, we propose an augmented Parallel-Pyramid Net (P^2Net) with an partial attention module. During data preprocessing, we proposed a differentiable auto data augmentation (DA^2) method in which sequences of data augmentations are formulated as a trainable and operational Convolution Neural Network (CNN) component. DA^2 improves the training efficiency and effectiveness. A parallel pyramid structure is followed to compensate the information loss introduced by the network. For the information loss problem in the backbone network, we optimize the backbone network by adopting a new parallel structure without increasing the overall computational complexity. To further refine the predictions after completion of global predictions, an Partial Attention Module (PAM) is defined to extract weighted features from different scale feature maps generated by the parallel pyramid structure. Compared with the traditional up-sampling refining, PAM can better capture the relationship between channels. Experiments corroborate the effectiveness of our proposed method. Notably, our method achieves the best performance on the challenging MSCOCO and MPII datasets.

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.

Recurrent Deep Attention Network for Person Re-Identification

Changhao Wang, Jun Zhou, Xianfei Duan, Guanwen Zhang, Wei Zhou

Auto-TLDR; Recurrent Deep Attention Network for Person Re-identification

Abstract Slides Poster Similar

Person re-identification (re-id) is an important task in video surveillance. It is challenging due to the appearance of person varying a wide range acrossnon-overlapping camera views. Recent years, attention-based models are introduced to learn discriminative representation. In this paper, we consider the attention selection in a natural way as like human moving attention on different parts of the visual field for person re-id. In concrete, we propose a Recurrent Deep Attention Network (RDAN) with an attention selection mechanism based on reinforcement learning. The RDAN aims to adaptively observe the identity-sensitive regions to build up the representation of individuals step by step. Extensive experiments on three person re-id benchmarks Market-1501, DukeMTMC-reID and CUHK03-NP demonstrate the proposed method can achieve competitive performance.

Learning with Delayed Feedback

Pranavan Theivendiram, Terence Sim

Auto-TLDR; Unsupervised Machine Learning with Delayed Feedback

Abstract Slides Poster Similar

We propose a novel supervised machine learning strategy, inspired by human learning, that enables an Agent to learn continually over its lifetime. A natural consequence is that the Agent must be able to handle an input whose label is delayed until a later time, or may not arrive at all. Our Agent learns in two steps: a short Seeding phase, in which the Agent's model is initialized with labelled inputs, and an indefinitely long Growing phase, in which the Agent refines and assesses its model if the label is given for an input, but stores the input in a finite-length queue if the label is missing. Queued items are matched against future input-label pairs that arrive, and the model is then updated. Our strategy also allows for the delayed feedback to take a different form. For example, in an image captioning task, the feedback could be a semantic segmentation rather than a textual caption. We show with many experiments that our strategy enables an Agent to learn flexibly and efficiently.

ILS-SUMM: Iterated Local Search for Unsupervised Video Summarization

Yair Shemer, Daniel Rotman, Nahum Shimkin

Auto-TLDR; ILS-SUMM: Iterated Local Search for Video Summarization

Abstract Slides Similar

In recent years, there has been an increasing interest in building video summarization tools, where the goal is to automatically create a short summary of an input video that properly represents the original content. We consider shot-based video summarization where the summary consists of a subset of the video shots which can be of various lengths. A straightforward approach to maximize the representativeness of a subset of shots is by minimizing the total distance between shots and their nearest selected shots. We formulate the task of video summarization as an optimization problem with a knapsack-like constraint on the total summary duration. Previous studies have proposed greedy algorithms to solve this problem approximately, but no experiments were presented to measure the ability of these methods to obtain solutions with low total distance. Indeed, our experiments on video summarization datasets show that the success of current methods in obtaining results with low total distance still has much room for improvement. In this paper, we develop ILS-SUMM, a novel video summarization algorithm to solve the subset selection problem under the knapsack constraint. Our algorithm is based on the well-known metaheuristic optimization framework -- Iterated Local Search (ILS), known for its ability to avoid weak local minima and obtain a good near-global minimum. Extensive experiments show that our method finds solutions with significantly better total distance than previous methods. Moreover, to indicate the high scalability of ILS-SUMM, we introduce a new dataset consisting of videos of various lengths.

Multiple Future Prediction Leveraging Synthetic Trajectories

Lorenzo Berlincioni, Federico Becattini, Lorenzo Seidenari, Alberto Del Bimbo

Auto-TLDR; Synthetic Trajectory Prediction using Markov Chains

Abstract Slides Poster Similar

Trajectory prediction is an important task, especially in autonomous driving. The ability to forecast the position of other moving agents can yield to an effective planning, ensuring safety for the autonomous vehicle as well for the observed entities. In this work we propose a data driven approach based on Markov Chains to generate synthetic trajectories, which are useful for training a multiple future trajectory predictor. The advantages are twofold: on the one hand synthetic samples can be used to augment existing datasets and train more effective predictors; on the other hand, it allows to generate samples with multiple ground truths, corresponding to diverse equally likely outcomes of the observed trajectory. We define a trajectory prediction model and a loss that explicitly address the multimodality of the problem and we show that combining synthetic and real data leads to prediction improvements, obtaining state of the art results.

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 Similar

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.

Adaptive Noise Injection for Training Stochastic Student Networks from Deterministic Teachers

Yi Xiang Marcus Tan, Yuval Elovici, Alexander Binder

Auto-TLDR; Adaptive Stochastic Networks for Adversarial Attacks

Abstract Slides Similar

Adversarial attacks have been a prevalent problem causing misclassification in machine learning models, with stochasticity being a promising direction towards greater robustness. However, stochastic networks frequently underperform compared to deterministic deep networks. In this work, we present a conceptually clear adaptive noise injection mechanism in combination with teacher-initialisation, which adjusts its degree of randomness dynamically through the computation of mini-batch statistics. This mechanism is embedded within a simple framework to obtain stochastic networks from existing deterministic networks. Our experiments show that our method is able to outperform prior baselines under white-box settings, exemplified through CIFAR-10 and CIFAR-100. Following which, we perform in-depth analysis on varying different components of training with our approach on the effects of robustness and accuracy, through the study of the evolution of decision boundary and trend curves of clean accuracy/attack success over differing degrees of stochasticity. We also shed light on the effects of adversarial training on a pre-trained network, through the lens of decision boundaries.

Spiking Neural Networks with Single-Spike Temporal-Coded Neurons for Network Intrusion Detection

Shibo Zhou, Xiaohua Li

Auto-TLDR; Spiking Neural Network with Leaky Neurons

Abstract Slides Poster Similar

Spiking neural network (SNN) is interesting due to its strong bio-plausibility and high energy efficiency. However, its performance is falling far behind conventional deep neural networks (DNNs). In this paper, considering a general class of single-spike temporal-coded integrate-and-fire neurons, we analyze the input-output expressions of both leaky and nonleaky neurons. We show that SNNs built with leaky neurons suffer from the overly-nonlinear and overly-complex input-output response, which is the major reason for their difficult training and low performance. This reason is more fundamental than the commonly believed problem of nondifferentiable spikes. To support this claim, we show that SNNs built with nonleaky neurons can have a less-complex and less-nonlinear input-output response. They can be easily trained and can have superior performance, which is demonstrated by experimenting with the SNNs over two popular network intrusion detection datasets, i.e., the NSL-KDD and the AWID datasets. Our experiment results show that the proposed SNNs outperform a comprehensive list of DNN models and classic machine learning models. This paper demonstrates that SNNs can be promising and competitive in contrast to common beliefs.

Active Sampling for Pairwise Comparisons via Approximate Message Passing and Information Gain Maximization

Aliaksei Mikhailiuk, Clifford Wilmot, Maria Perez-Ortiz, Dingcheng Yue, Rafal Mantiuk

Auto-TLDR; ASAP: An Active Sampling Algorithm for Pairwise Comparison Data

Abstract Slides Similar

Pairwise comparison data arise in many domains with subjective assessment experiments, for example in image and video quality assessment. In these experiments observers are asked to express a preference between two conditions. However, many pairwise comparison protocols require a large number of comparisons to infer accurate scores, which may be unfeasible when each comparison is time-consuming (e.g. videos) or expensive (e.g. medical imaging). This motivates the use of an active sampling algorithm that chooses only the most informative pairs for comparison. In this paper we propose ASAP, an active sampling algorithm based on approximate message passing and expected information gain maximization. Unlike most existing methods, which rely on partial updates of the posterior distribution, we are able to perform full updates and therefore much improve the accuracy of the inferred scores. The algorithm relies on three techniques for reducing computational cost: inference based on approximate message passing, selective evaluations of the information gain, and selecting pairs in a batch that forms a minimum spanning tree of the inverse of information gain. We demonstrate, with real and synthetic data, that ASAP offers the highest accuracy of inferred scores compared to the existing methods. We also provide an open-source GPU implementation of ASAP for large-scale experiments.

VOWEL: A Local Online Learning Rule for Recurrent Networks of Probabilistic Spiking Winner-Take-All Circuits

Hyeryung Jang, Nicolas Skatchkovsky, Osvaldo Simeone

Auto-TLDR; VOWEL: A Variational Online Local Training Rule for Winner-Take-All Spiking Neural Networks

Abstract Slides Similar

Networks of spiking neurons and Winner-Take-All spiking circuits (WTA-SNNs) can detect information encoded in spatio-temporal multi-valued events. These are described by the timing of events of interest, e.g., clicks, as well as by categorical numerical values assigned to each event, e.g., like or dislike. Other use cases include object recognition from data collected by neuromorphic cameras, which produce, for each pixel, signed bits at the times of sufficiently large brightness variations. Existing schemes for training WTA-SNNs are limited to rate-encoding solutions, and are hence able to detect only spatial patterns. Developing more general training algorithms for arbitrary WTA-SNNs inherits the challenges of training (binary) Spiking Neural Networks (SNNs). These amount, most notably, to the non-differentiability of threshold functions, to the recurrent behavior of spiking neural models, and to the difficulty of implementing backpropagation in neuromorphic hardware. In this paper, we develop a variational online local training rule for WTA-SNNs, referred to as VOWEL, that leverages only local pre- and post-synaptic information for visible circuits, and an additional common reward signal for hidden circuits. The method is based on probabilistic generalized linear neural models, control variates, and variational regularization. Experimental results on real-world neuromorphic datasets with multi-valued events demonstrate the advantages of WTA-SNNs over conventional binary SNNs trained with state-of-the-art methods, especially in the presence of limited computing resources.

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}.

Regularized Flexible Activation Function Combinations for Deep Neural Networks

Renlong Jie, Junbin Gao, Andrey Vasnev, Minh-Ngoc Tran

Auto-TLDR; Flexible Activation in Deep Neural Networks using ReLU and ELUs

Abstract Slides Poster Similar

Activation in deep neural networks is fundamental to achieving non-linear mappings. Traditional studies mainly focus on ﬁnding ﬁxed activations for a particular set of learning tasks or model architectures. The research on ﬂexible activation is quite limited in both designing philosophy and application scenarios. In this study, three principles of choosing ﬂexible activation components are proposed and a general combined form of ﬂexible activation functions is implemented. Based on this, a novel family of ﬂexible activation functions that can replace sigmoid or tanh in LSTM cells are implemented, as well as a new family by combining ReLU and ELUs. Also, two new regularisation terms based on assumptions as prior knowledge are introduced. It has been shown that LSTM models with proposed ﬂexible activations P-Sig-Ramp provide signiﬁcant improvements in time series forecasting, while the proposed P-E2-ReLU achieves better and more stable performance on lossy image compression tasks with convolutional auto-encoders. In addition, the proposed regularization terms improve the convergence,performance and stability of the models with ﬂexible activation functions. The code for this paper is available at https://github.com/9NXJRDDRQK/Flexible Activation.

Stage-Wise Neural Architecture Search

Artur Jordão, Fernando Akio Yamada, Maiko Lie, William Schwartz

Auto-TLDR; Efficient Neural Architecture Search for Deep Convolutional Networks

Abstract Slides Poster Similar

Modern convolutional networks such as ResNet and NASNet have achieved state-of-the-art results in many computer vision applications. These architectures consist of stages, which are sets of layers that operate on representations in the same resolution. It has been demonstrated that increasing the number of layers in each stage improves the prediction ability of the network. However, the resulting architecture becomes computationally expensive in terms of floating point operations, memory requirements and inference time. Thus, significant human effort is necessary to evaluate different trade-offs between depth and performance. To handle this problem, recent works have proposed to automatically design high-performance architectures, mainly by means of neural architecture search (NAS). Current NAS strategies analyze a large set of possible candidate architectures and, hence, require vast computational resources and take many GPUs days. Motivated by this, we propose a NAS approach to efficiently design accurate and low-cost convolutional architectures and demonstrate that an efficient strategy for designing these architectures is to learn the depth stage-by-stage. For this purpose, our approach increases depth incrementally in each stage taking into account its importance, such that stages with low importance are kept shallow while stages with high importance become deeper. We conduct experiments on the CIFAR and different versions of ImageNet datasets, where we show that architectures discovered by our approach achieve better accuracy and efficiency than human-designed architectures. Additionally, we show that architectures discovered on CIFAR-10 can be successfully transferred to large datasets. Compared to previous NAS approaches, our method is substantially more efficient, as it evaluates one order of magnitude fewer models and yields architectures on par with the state-of-the-art.