Learning disentangled representations of natural language is essential for many NLP tasks, e.g., conditional text generation, style transfer, personalized dialogue systems, etc. Similar problems have been studied extensively for other forms of data, such as images and videos. However, the discrete nature of natural language makes the disentangling of textual representations more challenging (e.g., the manipulation over the data space cannot be easily achieved). Inspired by information theory, we propose a novel method that effectively manifests disentangled representations of text, without any supervision on semantics. A new mutual information upper bound is derived and leveraged to measure dependence between style and content. By minimizing this upper bound, the proposed method induces style and content embeddings into two independent low-dimensional spaces. Experiments on both conditional text generation and text-style transfer demonstrate the high quality of our disentangled representation in terms of content and style preservation.

The signature transform is a 'universal nonlinearity' on the space of continuous vector-valued paths, and has received attention for use in machine learning on time series. However, real-world temporal data is typically observed at discrete points in time, and must first be transformed into a continuous path before signature techniques can be applied. We make this step explicit by characterising it as an imputation problem, and empirically assess the impact of various imputation strategies when applying signature-based neural nets to irregular time series data. For one of these strategies, Gaussian process (GP) adapters, we propose an extension~(GP-PoM) that makes uncertainty information directly available to the subsequent classifier while at the same time preventing costly Monte-Carlo (MC) sampling. In our experiments, we find that the choice of imputation drastically affects shallow signature models, whereas deeper architectures are more robust. Next, we observe that uncertainty-aware predictions (based on GP-PoM or indicator imputations) are beneficial for predictive performance, even compared to the uncertainty-aware training of conventional GP adapters. In conclusion, we have demonstrated that the path construction is indeed crucial for signature models and that our proposed strategy leads to competitive performance in general, while improving robustness of signature models in particular.

This is the second installment in a series. The ultimate root of the stupidity of AI systems, I argue, lies in their strictly algorithmic character. AI as presently understood is based on digital processing systems that carry out binary-numerical operations in a step-by-step fashion according to fixed sets of algorithms, starting from an array of numerical inputs. Some may object to this characterization, pointing out that AI systems can constantly change their own "rules" – reprogramming themselves, so to speak. That is true; but the self-reprogramming must follow some algorithm.

A team of researchers from the Chinese Academy of Sciences and the City University of Hong Kong has introduced a local-to-global approach that can generate lifelike human portraits from relatively rudimentary sketches. Recent deep image-to-image translation techniques have enabled the prompt generation of human face images from sketches, but these methods tend to suffer from overfitting to their inputs. They thus achieve the most realistic results only when the source drawings have high-quality artistry or are accompanied by edge maps. Unlike most deep learning based solutions for sketch-to-image translation that take input sketches as fixed, 'hard' constraints and then attempt to reconstruct the missing texture or shading information between strokes, the key idea behind the new approach is to implicitly learn a space of plausible face sketches from real face sketch images and find the point in this space that best approximates the input sketch. Because this approach treats input sketches more as'soft' constraints that will guide image synthesis, it is able to produce high-quality face images with increased plausibility even from rough and/or incomplete inputs.

Plug-and-play priors (PnP) is a broadly applicable methodology for solving inverse problems by exploiting statistical priors specified as denoisers. Recent work has reported the state-of-the-art performance of PnP algorithms using pre-trained deep neural nets as denoisers in a number of imaging applications. However, current PnP algorithms are impractical in large-scale settings due to their heavy computational and memory requirements. This work addresses this issue by proposing an incremental variant of the widely used PnP-ADMM algorithm, making it scalable to large-scale datasets. We theoretically analyze the convergence of the algorithm under a set of explicit assumptions, extending recent theoretical results in the area. Additionally, we show the effectiveness of our algorithm with nonsmooth data-fidelity terms and deep neural net priors, its fast convergence compared to existing PnP algorithms, and its scalability in terms of speed and memory.

The identification of nonlinear dynamics from observations is essential for the alignment of the theoretical ideas and experimental data. The last, in turn, is often corrupted by the side effects and noise of different natures, so probabilistic approaches could give a more general picture of the process. At the same time, high-dimensional probabilities modeling is a challenging and data-intensive task. In this paper, we establish a parallel between the dynamical systems modeling and language modeling tasks. We propose a transformer-based model that incorporates geometrical properties of the data and provide an iterative training algorithm allowing the fine-grid approximation of the conditional probabilities of high-dimensional dynamical systems.

Bilevel optimization (BLO) is a popular approach with many applications including hyperparameter optimization, neural architecture search, adversarial robustness and model-agnostic meta-learning. However, the approach suffers from time and memory complexity proportional to the length $r$ of its inner optimization loop, which has led to several modifications being proposed. One such modification is \textit{first-order} BLO (FO-BLO) which approximates outer-level gradients by zeroing out second derivative terms, yielding significant speed gains and requiring only constant memory as $r$ varies. Despite FO-BLO's popularity, there is a lack of theoretical understanding of its convergence properties. We make progress by demonstrating a rich family of examples where FO-BLO-based stochastic optimization does not converge to a stationary point of the BLO objective. We address this concern by proposing a new FO-BLO-based unbiased estimate of outer-level gradients, enabling us to theoretically guarantee this convergence, with no harm to memory and expected time complexity. Our findings are supported by experimental results on Omniglot and Mini-ImageNet, popular few-shot meta-learning benchmarks.

We introduce the problem of explaining graph generation, formulated as controlling the generative process to produce graphs of explainable desired structures. By directing this generative process, we can measure and explain the observed outcomes. We propose SHADOWCAST, a controllable generative model capable of mimicking real-world networks and directing the generation, as a novel approach to this problem. The proposed model is based on a conditional generative adversarial network for graph data. We design it with the capability to maintain generation control by managing the conditions. Comprehensive experiments on three real-world network datasets demonstrate our model's competitive performance in the graph generation task. Furthermore, we direct SHADOWCAST to generate explainable graphs of different structures to show its effective controllability. As the first work to pose the problem of explaining generated graphs by controlling the generation, SHADOWCAST paves the way for future research in this exciting area.

When thrust into an unfamiliar environment and charged with solving a series of tasks, an effective agent should (1) leverage prior knowledge to solve its current task while (2) efficiently exploring to gather knowledge for use in future tasks, and then (3) plan using that knowledge when faced with new tasks in that same environment. We introduce two domains for conducting research on this challenge, and find that state-of-the-art deep reinforcement learning (RL) agents fail to plan in novel environments. We develop a recursive implicit planning module that operates over episodic memories, and show that the resulting deep-RL agent is able to explore and plan in novel environments, outperforming the nearest baseline by factors of 2-3 across the two domains. We find evidence that our module (1) learned to execute a sensible information-propagating algorithm and (2) generalizes to situations beyond its training experience.

Many unsupervised representation learning methods belong to the class of similarity learning models. While various modality-specific approaches exist for different types of data, a core property of many methods is that representations of similar inputs are close under some similarity function. We propose EMDE (Efficient Manifold Density Estimator) - a framework utilizing arbitrary vector representations with the property of local similarity to succinctly represent smooth probability densities on Riemannian manifolds. Our approximate representation has the desirable properties of being fixed-size and having simple additive compositionality, thus being especially amenable to treatment with neural networks - both as input and output format, producing efficient conditional estimators. We generalize and reformulate the problem of multi-modal recommendations as conditional, weighted density estimation on manifolds. Our approach allows for trivial inclusion of multiple interaction types, modalities of data as well as interaction strengths for any recommendation setting. Applying EMDE to both top-k and session-based recommendation settings, we establish new state-of-the-art results on multiple open datasets in both uni-modal and multi-modal settings. We release the source code and our own real-world dataset of e-commerce product purchases, with special focus on modeling of the item cold-start problem.