In this work we aimed at comparing our findings in depression detection task with methodologies applied in present literature. Previously we showed that when electrophysiological signal (in this case electroencephalogram, EEG) is characterized by nonlinear measures, any of seven most popular classifiers yields high accuracy on the task. Following every step we done in this process we compare it with other researchers' practice and comment on other findings mainly from analysis of electrical signals or nonlinear analysis showing what would be optimal for further research. We focused on discussing various mistakes and differences that could potentially lead to unwarranted optimism and other misinterpretation of results. In Conclusion we summarize recommendation for future research in order to be applicable in clinical practice. Introduction Current clinical psychiatry is lacking objective biochemical or electrophysiological tests used for diagnosis unlike other medical disciplines. To diagnose depression, clinician will typically rely on the self-report from the patient and his experience in applying DSM manual, which is standardized list of symptoms to be checked in every case (in order to be qualified as a certain disorder). It is perfectly possible that two persons diagnosed with the same disorder have not overlapping symptoms, and that one person can have two distinct diagnosis. If someone has more than three episodes of depression, that is considered to be recurrent depression (after every episode the probability of the next one is doubling). This is particularly heard to treat and manage therapy which is ongoing through person's whole life. Apart from obsolete diagnostic, all antidepressants have serious side-effects, the waiting lists are very long (in Nederland they are between 6 and 9 months long) and the therapy can last for years or even decades. It is reported than only 11 - 30% of patients are improving in the first year of therapy (Rush et al., 2008).

Distributed parallel stochastic gradient descent algorithms are workhorses for large scale machine learning tasks. Among them, local stochastic gradient descent (Local SGD) has attracted significant attention due to its low communication complexity. Previous studies prove that the communication complexity of Local SGD with a fixed or an adaptive communication period is in the order of $O (N^{\frac{3}{2}} T^{\frac{1}{2}})$ and $O (N^{\frac{3}{4}} T^{\frac{3}{4}})$ when the data distributions on clients are identical (IID) or otherwise (Non-IID). In this paper, to accelerate the convergence by reducing the communication complexity, we propose \textit{ST}agewise \textit{L}ocal \textit{SGD} (STL-SGD), which increases the communication period gradually along with decreasing learning rate. We prove that STL-SGD can keep the same convergence rate and linear speedup as mini-batch SGD. In addition, as the benefit of increasing the communication period, when the objective is strongly convex or satisfies the Polyak-\L ojasiewicz condition, the communication complexity of STL-SGD is $O (N \log{T})$ and $O (N^{\frac{1}{2}} T^{\frac{1}{2}})$ for the IID case and the Non-IID case respectively, achieving significant improvements over Local SGD. Experiments on both convex and non-convex problems demonstrate the superior performance of STL-SGD.

Thompson sampling has become a ubiquitous approach to online decision problems with bandit feedback. The key algorithmic task for Thompson sampling is drawing a sample from the posterior of the optimal action. We propose an alternative arm selection rule we dub TS-UCB, that requires negligible additional computational effort but provides significant performance improvements relative to Thompson sampling. At each step, TS-UCB computes a score for each arm using two ingredients: posterior sample(s) and upper confidence bounds. TS-UCB can be used in any setting where these two quantities are available, and it is flexible in the number of posterior samples it takes as input. This proves particularly valuable in heuristics for deep contextual bandits: we show that TS-UCB achieves materially lower regret on all problem instances in a deep bandit suite proposed in Riquelme et al. (2018). Finally, from a theoretical perspective, we establish optimal regret guarantees for TS-UCB for both the K-armed and linear bandit models.

In this paper, we discuss application of iterative Stochastic Optimization routines to the problem of sparse signal recovery from noisy observation. Using Stochastic Mirror Descent algorithm as a building block, we develop a multistage procedure for recovery of sparse solutions to Stochastic Optimization problem under assumption of smoothness and quadratic minoration on the expected objective. An interesting feature of the proposed algorithm is its linear convergence of the approximate solution during the preliminary phase of the routine when the component of stochastic error in the gradient observation which is due to bad initial approximation of the optimal solution is larger than the "ideal" asymptotic error component owing to observation noise "at the optimal solution." We also show how one can straightforwardly enhance reliability of the corresponding solution by using Median-of-Means like techniques. We illustrate the performance of the proposed algorithms in application to classical problems of recovery of sparse and low rank signals in linear regression framework. We show, under rather weak assumption on the regressor and noise distributions, how they lead to parameter estimates which obey (up to factors which are logarithmic in problem dimension and confidence level) the best known to us accuracy bounds.

The problem of Catastrophic Forgetting has received a lot of attention in the past years. An important class of proposed solutions are so-called regularisation approaches, which protect weights from large changes according to their importances. Various ways to measure this importance have been put forward, all stemming from different theoretical or intuitive motivations. We present mathematical and empirical evidence that two of these methods -- Synaptic Intelligence and Memory Aware Synapses -- approximate a rescaled version of the Fisher Information, a theoretically justified importance measure also used in the literature. As part of our methods, we show that the importance approximation of Synaptic Intelligence is biased and that, in fact, this bias explains its performance best. Altogether, our results offer a theoretical account for the effectiveness of different regularisation approaches and uncover similarities between the methods proposed so far.

In this article, we propose a new variational approach to learn private and/or fair representations. This approach is based on the Lagrangians of a new formulation of the privacy and fairness optimization problems that we propose. In this formulation, we aim at generating representations of the data that keep a prescribed level of the relevant information that is not shared by the private or sensitive data, while minimizing the remaining information they keep. The proposed approach (i) exhibits the similarities of the privacy and fairness problems, (ii) allows us to control the trade-off between utility and privacy or fairness through the Lagrange multiplier parameter, and (iii) can be comfortably incorporated to common representation learning algorithms such as the VAE, the $\beta$-VAE, the VIB, or the nonlinear IB.

Stochastic optimization is the process of minimizing a deterministic objective function via the simulation of random elements, and it is one of the most successful methods for optimizing complex or unknown objectives. Relatively simple stochastic optimization procedures--in particular, stochastic gradient descent (SGD)--have become the backbone of modern machine learning (ML) [50]. To improve understanding of stochastic optimization in ML, and particularly why SGD and its extensions work so well, recent theoretical work has sought to study its properties and dynamics [47]. Such analyses typically approach the problem through one of two perspectives. The first perspective, an optimization (or quenching) perspective, examines convergence either in expectation [11, 20, 28, 60, 84] or with some positive (high) probability [19, 41, 66, 77] through the lens of a deterministic counterpart. This perspective inherits some limitations of deterministic optimizers, including assumptions (e.g., convexity, Polyak-Łojasiewicz criterion, etc.) that are either not satisfied by state-of-the-art problems, or not strong enough to imply convergence to a quality (e.g., global) optimum. More concerning, however, is the inability to explain what has come to be known as the "generalization gap" phenomenon: increasing stochasticity by reducing batch size appears to improve generalization performance [38, 55]. Empirically, existing strategies do tend to break down for inference tasks when using large batch sizes [27].

Although variational autoencoders (VAE) are successfully used to obtain meaningful low-dimensional representations for high-dimensional data, aspects of their loss function are not yet fully understood. We introduce a theoretical framework that is based on a connection between VAE and generalized linear models (GLM). The equality between the activation function of a VAE and the inverse of the link function of a GLM enables us to provide a systematic generalization of the loss analysis for VAE based on the assumption that the distribution of the decoder belongs to an exponential dispersion family (EDF). As a further result, we can initialize VAE nets by maximum likelihood estimates (MLE) that enhance the training performance on both synthetic and real world data sets.

In real-world classification problems, pairwise supervision (i.e., a pair of patterns with a binary label indicating whether they belong to the same class or not) can often be obtained at a lower cost than ordinary class labels. Similarity learning is a general framework to utilize such pairwise supervision to elicit useful representations by inferring the relationship between two data points, which encompasses various important preprocessing tasks such as metric learning, kernel learning, graph embedding, and contrastive representation learning. Although elicited representations are expected to perform well in downstream tasks such as classification, little theoretical insight has been given in the literature so far. In this paper, we reveal that a specific formulation of similarity learning is strongly related to the objective of binary classification, which spurs us to learn a binary classifier without ordinary class labels---by fitting the product of real-valued prediction functions of pairwise patterns to their similarity. Our formulation of similarity learning does not only generalize many existing ones, but also admits an excess risk bound showing an explicit connection to classification. Finally, we empirically demonstrate the practical usefulness of the proposed method on benchmark datasets.

Exploring the transition dynamics is essential to the success of reinforcement learning (RL) algorithms. To face the challenges of exploration, we consider a zero-shot meta RL framework that completely separates exploration from exploitation and is suitable for the meta RL setting where there are many reward functions of interest. In the exploration phase, the agent learns an exploratory policy by interacting with a reward-free environment and collects a dataset of transitions by executing the policy. In the planning phase, the agent computes a good policy for any reward function based on the dataset without further interacting with the environment. This framework brings new challenges for exploration algorithms. In the exploration phase, we propose to maximize the R\'enyi entropy over the state-action space and justify this objective theoretically. We further deduce a policy gradient formulation for this objective and design a practical exploration algorithm that can deal with complex environments based on PPO. In the planning phase, we use a batch RL algorithm, batch constrained deep Q-learning (BCQ), to solve for good policies given arbitrary reward functions. Empirically, we show that our exploration algorithm is effective and sample efficient, and results in superior policies for arbitrary reward functions in the planning phase.