Variational data assimilation and deep learning share many algorithmic aspects in common. While the former focuses on system state estimation, the latter provides great inductive biases to learn complex relationships. We here design a hybrid architecture learning the assimilation task directly from partial and noisy observations, using the mechanistic constraint of the 4DVAR algorithm. Finally, we show in an experiment that the proposed method was able to learn the desired inversion with interesting regularizing properties and that it also has computational interests.

Given an environment (e.g., a simulator) for evaluating samples in a specified design space and a set of weighted evaluation metrics -- one can use Theta-Resonance, a single-step Markov Decision Process (MDP), to train an intelligent agent producing progressively more optimal samples. In Theta-Resonance, a neural network consumes a constant input tensor and produces a policy as a set of conditional probability density functions (PDFs) for sampling each design dimension. We specialize existing policy gradient algorithms in deep reinforcement learning (D-RL) in order to use evaluation feedback (in terms of cost, penalty or reward) to update our policy network with robust algorithmic stability and minimal design evaluations. We study multiple neural architectures (for our policy network) within the context of a simple SoC design space and propose a method of constructing synthetic space exploration problems to compare and improve design space exploration (DSE) algorithms. Although we only present categorical design spaces, we also outline how to use Theta-Resonance in order to explore continuous and mixed continuous-discrete design spaces.

Uncertainty plays a crucial role in the machine learning field. Both model trustworthiness and performance require the understanding of uncertainty, especially for models used in high-stake applications where errors can cause cataclysmic consequences, such as medical diagnosis and autonomous driving. Accordingly, uncertainty decomposition and quantification have attracted more and more attention in recent years. This short report aims to demystify the notion of uncertainty decomposition through an introduction to two types of uncertainty and several decomposition exemplars, including maximum likelihood estimation, Gaussian processes, deep neural network, and ensemble learning. In the end, cross connections to other topics in this seminar and two conclusions are provided.

The study of quasar light curves poses two problems: inference of the power spectrum and interpolation of an irregularly sampled time series. A baseline approach to these tasks is to interpolate a time series with a Damped Random Walk (DRW) model, in which the spectrum is inferred using Maximum Likelihood Estimation (MLE). However, the DRW model does not describe the smoothness of the time series, and MLE faces many problems in terms of optimization and numerical precision. In this work, we introduce a new stochastic model that we call $\textit{Convolved Damped Random Walk}$ (CDRW). This model introduces a concept of smoothness to a DRW, which enables it to describe quasar spectra completely. We also introduce a new method of inference of Gaussian process parameters, which we call $\textit{Neural Inference}$. This method uses the powers of state-of-the-art neural networks to improve the conventional MLE inference technique. In our experiments, the Neural Inference method results in significant improvement over the baseline MLE (RMSE: $0.318 \rightarrow 0.205$, $0.464 \rightarrow 0.444$). Moreover, the combination of both the CDRW model and Neural Inference significantly outperforms the baseline DRW and MLE in interpolating a typical quasar light curve ($\chi^2$: $0.333 \rightarrow 0.998$, $2.695 \rightarrow 0.981$). The code is published on GitHub.

Clustering is one of the most important unsupervised learning techniques for understanding the underlying data structures. The goal is to partition a data set into many subsets, called clusters, such that the observations within the subsets are the most homogeneous and the observations between the subsets are the most heterogeneous. Clustering is usually carried out by specifying a similarity or dissimilarity measure between observations. Examples include the k-means [17, 19, 29, 37], the k-medians [3], the k-modes [5], and the generalized k-means [2, 31, 45], as well as many of their modifications [21, 24, 42]. Among those, the k-means has been considered as one of the most straightforward and popular methods since it was proposed sixty years ago [23, 36]. Although it is well known, the investigation of the theoretical properties is still far behind, leading to difficulties in developing more precise k-means methods in practice. The goal of the present research is to propose a new concept called clustering consistency for the asymptotics of the k-means with a resulting clustering method better than the existing k-means methods adopted by many software packages, including those adopted by R and Python.

We propose a decoding-based approach to detect context effects on neural codes in longitudinal neural recording data. The approach is agnostic to how information is encoded in neural activity, and can control for a variety of possible confounding factors present in the data. We demonstrate our approach by determining whether it is possible to decode location encoding from prefrontal cortex in the mouse and, further, testing whether the encoding changes due to task engagement.

Active learning has been studied extensively as a method for efficient data collection. Among the many approaches in literature, Expected Error Reduction (EER) (Roy and McCallum) has been shown to be an effective method for active learning: select the candidate sample that, in expectation, maximally decreases the error on an unlabeled set. However, EER requires the model to be retrained for every candidate sample and thus has not been widely used for modern deep neural networks due to this large computational cost. In this paper we reformulate EER under the lens of Bayesian active learning and derive a computationally efficient version that can use any Bayesian parameter sampling method (such as arXiv:1506.02142). We then compare the empirical performance of our method using Monte Carlo dropout for parameter sampling against state of the art methods in the deep active learning literature. Experiments are performed on four standard benchmark datasets and three WILDS datasets (arXiv:2012.07421). The results indicate that our method outperforms all other methods except one in the data shift scenario: a model dependent, non-information theoretic method that requires an order of magnitude higher computational cost (arXiv:1906.03671).

Stuttering is a speech disorder during which the flow of speech is interrupted by involuntary pauses and repetition of sounds. Stuttering identification is an interesting interdisciplinary domain research problem which involves pathology, psychology, acoustics, and signal processing that makes it hard and complicated to detect. Recent developments in machine and deep learning have dramatically revolutionized speech domain, however minimal attention has been given to stuttering identification. This work fills the gap by trying to bring researchers together from interdisciplinary fields. In this paper, we review comprehensively acoustic features, statistical and deep learning based stuttering/disfluency classification methods. We also present several challenges and possible future directions.

Machine Learning involves algorithms that learn from patterns of data and then apply it to decision making. Deep Learning, on the other hand, is able to learn through processing data on its own and is quite similar to the human brain where it identifies something, analyse it, and makes a decision.

Sampling-based methods are widely adopted solutions for robot motion planning. The methods are straightforward to implement, effective in practice for many robotic systems. It is often possible to prove that they have desirable properties, such as probabilistic completeness and asymptotic optimality. Nevertheless, they still face challenges as the complexity of the underlying planning problem increases, especially under tight computation time constraints, which impact the quality of returned solutions or given inaccurate models. This has motivated machine learning to improve the computational efficiency and applicability of Sampling-Based Motion Planners (SBMPs). This survey reviews such integrative efforts and aims to provide a classification of the alternative directions that have been explored in the literature. It first discusses how learning has been used to enhance key components of SBMPs, such as node sampling, collision detection, distance or nearest neighbor computation, local planning, and termination conditions. Then, it highlights planners that use learning to adaptively select between different implementations of such primitives in response to the underlying problem's features. It also covers emerging methods, which build complete machine learning pipelines that reflect the traditional structure of SBMPs. It also discusses how machine learning has been used to provide data-driven models of robots, which can then be used by a SBMP. Finally, it provides a comparative discussion of the advantages and disadvantages of the approaches covered, and insights on possible future directions of research. An online version of this survey can be found at: https://prx-kinodynamic.github.io/