Generative adversarial networks (GANs) used in domain adaptation tasks have the ability to generate images that are both realistic and personalized, transforming an input image while maintaining its identifiable characteristics. However, they often require a large quantity of training data to produce high-quality images in a robust way, which limits their usability in cases when access to data is limited. In our paper, we explore the potential of using images from a simulated 3D environment to improve a domain adaptation task carried out by the MUNIT architecture, aiming to use the resulting images to raise awareness of the potential future impacts of climate change.
Despite the growing interest in unsupervised learning, extracting meaningful knowledge from unlabelled audio remains an open challenge. To take a step in this direction, we recently proposed a problem-agnostic speech encoder (PASE), that combines a convolutional encoder followed by multiple neural networks, called workers, tasked to solve self-supervised problems (i.e., ones that do not require manual annotations as ground truth). PASE was shown to capture relevant speech information, including speaker voice-print and phonemes. This paper proposes PASE+, an improved version of PASE for robust speech recognition in noisy and reverberant environments. To this end, we employ an online speech distortion module, that contaminates the input signals with a variety of random disturbances. We then propose a revised encoder that better learns short- and long-term speech dynamics with an efficient combination of recurrent and convolutional networks. Finally, we refine the set of workers used in self-supervision to encourage better cooperation. Results on TIMIT, DIRHA and CHiME-5 show that PASE+ significantly outperforms both the previous version of PASE as well as common acoustic features. Interestingly, PASE+ learns transferable representations suitable for highly mismatched acoustic conditions.
Transfer in Reinforcement Learning (RL) refers to the idea of applying knowledge gained from previous tasks to solving related tasks. Learning a universal value function (Schaul et al., 2015), which generalizes over goals and states, has previously been shown to be useful for transfer. However, successor features are believed to be more suitable than values for transfer (Dayan, 1993; Barreto et al.,2017), even though they cannot directly generalize to new goals. In this paper, we propose (1) Universal Successor Features (USFs) to capture the underlying dynamics of the environment while allowing generalization to unseen goals and (2) a flexible end-to-end model of USFs that can be trained by interacting with the environment. We show that learning USFs is compatible with any RL algorithm that learns state values using a temporal difference method. Our experiments in a simple gridworld and with two MuJoCo environments show that USFs can greatly accelerate training when learning multiple tasks and can effectively transfer knowledge to new tasks.
There is an analogy between machine learning systems and economic entities in that they are both adaptive, and their behaviour is specified in a more-or-less explicit way. It appears that the area of AI that is most analogous to the behaviour of economic entities is that of morally good decision-making, but it is an open question as to how precisely moral behaviour can be achieved in an AI system. This paper explores the analogy between these two complex systems, and we suggest that a clearer understanding of this apparent analogy may help us forward in both the socio-economic domain and the AI domain: known results in economics may help inform feasible solutions in AI safety, but also known results in AI may inform economic policy. If this claim is correct, then the recent successes of deep learning for AI suggest that more implicit specifications work better than explicit ones for solving such problems.
Much of the existing research on the social and ethical impact of Artificial Intelligence has been focused on defining ethical principles and guidelines surrounding Machine Learning (ML) and other Artificial Intelligence (AI) algorithms [IEEE, 2017, Jobin et al., 2019]. While this is extremely useful for helping define the appropriate social norms of AI, we believe that it is equally important to discuss both the potential and risks of ML and to inspire the community to use ML for beneficial objectives. In the present article, which is specifically aimed at ML practitioners, we thus focus more on the latter, carrying out an overview of existing high-level ethical frameworks and guidelines, but above all proposing both conceptual and practical principles and guidelines for ML research and deployment, insisting on concrete actions that can be taken by practitioners to pursue a more ethical and moral practice of ML aimed at using AI for social good.
We propose a novel approach using supervised learning to obtain near-optimal primal solutions for two-stage stochastic integer programming (2SIP) problems with constraints in the first and second stages. The goal of the algorithm is to predict a "representative scenario" (RS) for the problem such that, deterministically solving the 2SIP with the random realization equal to the RS, gives a near-optimal solution to the original 2SIP. Predicting an RS, instead of directly predicting a solution ensures first-stage feasibility of the solution. If the problem is known to have complete recourse, second-stage feasibility is also guaranteed. For computational testing, we learn to find an RS for a two-stage stochastic facility location problem with integer variables and linear constraints in both stages and consistently provide near-optimal solutions. Our computing times are very competitive with those of general-purpose integer programming solvers to achieve a similar solution quality.
In this paper, we propose a Generative Translation Classification Network (GTCN) for improving visual classification accuracy in settings where classes are visually similar and data is scarce. For this purpose, we propose joint learning to train a classifier and a generative stochastic translation network end-to-end. The translation network is used to perform on-line data augmentation across classes, whereas previous works have mostly involved domain adaptation. To help the model further benefit from this data-augmentation, we introduce an adaptive fade-in loss and a quadruplet loss. We perform experiments on multiple datasets to demonstrate the proposed method's performance in varied settings. Of particular interest, training on 40% of the dataset is enough for our model to surpass the performance of baselines trained on the full dataset. When our architecture is trained on the full dataset, we achieve comparable performance with state-of-the-art methods despite using a light-weight architecture.
The CLEVR dataset of natural-looking questions about 3D-rendered scenes has recently received much attention from the research community. A number of models have been proposed for this task, many of which achieved very high accuracies of around 97-99%. In this work, we study how systematic the generalization of such models is, that is to which extent they are capable of handling novel combinations of known linguistic constructs. To this end, we test models' understanding of referring expressions based on matching object properties (such as e.g. "the object that is the same size as the red ball") in novel contexts. Our experiments on the thereby constructed CLOSURE benchmark show that state-of-the-art models often do not exhibit systematicity after being trained on CLEVR. Surprisingly, we find that an explicitly compositional Neural Module Network model also generalizes badly on CLOSURE, even when it has access to the ground-truth programs at test time. We improve the NMN's systematic generalization by developing a novel Vector-NMN module architecture with vector-valued inputs and outputs. Lastly, we investigate the extent to which few-shot transfer learning can help models that are pretrained on CLEVR to adapt to CLOSURE. Our few-shot learning experiments contrast the adaptation behavior of the models with intermediate discrete programs with that of the end-to-end continuous models.
Centered Kernel Alignment (CKA) was recently proposed as a similarity metric for comparing activation patterns in deep networks. Here we experiment with the modified RV-coefficient (RV2), which has very similar properties as CKA while being less sensitive to dataset size. We compare the representations of networks that received varying amounts of training on different layers: a standard trained network (all parameters updated at every step), a freeze trained network (layers gradually frozen during training), random networks (only some layers trained), and a completely untrained network. We found that RV2 was able to recover expected similarity patterns and provide interpretable similarity matrices that suggested hypotheses about how representations are affected by different training recipes. We propose that the superior performance achieved by freeze training can be attributed to representational differences in the penultimate layer. Our comparisons of random networks suggest that the inputs and targets serve as anchors on the representations in the lowest and highest layers.
In this work, we present the application of convolutional neural networks for segmenting water bodies in satellite images. We first use a variant of the U-Net model to segment rivers and lakes from very high-resolution images from Peru. To circumvent the issue of scarce labelled data, we investigate the applicability of a knowledge transfer-based model that learns the mapping from high-resolution labelled images and combines it with the very high-resolution mapping so that better segmentation can be achieved. We train this model in a single process, end-to-end. Our preliminary results show that adding the information from the available high-resolution images does not help out-of-the-box, and in fact worsen results. This leads us to infer that the high-resolution data could be from a different distribution, and its addition leads to increased variance in our results.