Abstract--We address the problem of assisting human dispatchers in operating power grids in today's changing context using machine learning, with the aim of increasing security and reducing costs. Power networks are highly regulated systems, which at all times must meet varying demands of electricity with a complex production system, including conventional power plants, less predictable renewable energies (such as wind or solar power), and the possibility of buying/selling electricity on the international market with more and more actors involved at a European scale. This problem is becoming ever more challenging in an aging network infrastructure. One of the primary goals of dispatchers is to protect equipment (e.g. Using years of historical data collected by the French Transmission Service Operator (TSO) "Réseau de Transport d'Electricité" (RTE), we develop novel machine learning techniques (drawing on "deep learning") to mimic human decisions to devise "remedial actions" to prevent any line to violate power flow limits (so-called "thermal limits"). The proposed technique is hybrid. It does not rely purely on machine learning: every action will be tested with actual simulators before being proposed to the dispatchers or implemented on the grid. Electricity is a commodity that consumers take for granted and, while governments relaying public opinion (rightfully) request that renewable energies be used increasingly, little is known about what this entails behind the scenes in additional complexity for the Transmission Service Operators (TSOs) to operate the power grid in security. Indeed, renewable energies such as wind and solar power are less predictable than conventional power sources (mainly thermal power plants).

Generative modeling, a typical unsupervised learning that makes use of huge amount of unlabeled data, lies in the heart of rapid development of modern machine learning techniques [1]. Different from discriminative tasks such as pattern recognition, the goal of generative modeling is to model the probability distribution of input data and thus be able to generate new samples according to the distribution. At the research frontier of generative modeling, it was used for finding good data representation and dealing with tasks with missing data. Popular generative machine learning models include the Boltzmann Machines (BM) [2, 3] and their generalizations [4], variational autoencoders (VAE) [5], autoregressive models [6, 7], nonlinear density estimations [8-10], and the generative adversarial networks (GAN) [11]. For generative model design, one tries to balance the representational power and efficiency of learning and sampling. There is a long history of relation between generative modeling and physics, especially statistical physics. Some celebrated models, such as Hopfield model [12], and Boltzmann machine [2, 3], are closely related to the Ising model in statistical physics, and its inverse version which learns couplings in the Ising model based on given training configurations [13, 14]. The task of generative modeling also shares many similarities with quantum physics research in the sense that both of them try to model probability distributions in an enormously large space. In the past decades, tensor network (TN) states and algorithms have been shown to be an incredibly potent tool set for studying many-body quantum physics with its power in expressing quantum states relevant to realistic situations [15, 16].

We propose a Bayesian regression method that accounts for multi-way interactions of arbitrary orders among the predictor variables. Our model makes use of a factorization mechanism for representing the regression coefficients of interactions among the predictors, while the interaction selection is guided by a prior distribution on random hypergraphs, a construction which generalizes the Finite Feature Model. We present a posterior inference algorithm based on Gibbs sampling, and establish posterior consistency of our regression model. Our method is evaluated with extensive experiments on simulated data and demonstrated to be able to identify meaningful interactions in applications in genetics and retail demand forecasting.

This paper presents a technique for reduced-order Markov modeling for compact representation of time-series data. In this work, symbolic dynamics-based tools have been used to infer an approximate generative Markov model. The time-series data are first symbolized by partitioning the continuous measurement space of the signal and then, the discrete sequential data are modeled using symbolic dynamics. In the proposed approach, the size of temporal memory of the symbol sequence is estimated from spectral properties of the resulting stochastic matrix corresponding to a first-order Markov model of the symbol sequence. Then, hierarchical clustering is used to represent the states of the corresponding full-state Markov model to construct a reduced-order or size Markov model with a non-deterministic algebraic structure. Subsequently, the parameters of the reduced-order Markov model are identified from the original model by making use of a Bayesian inference rule. The final model is selected using information-theoretic criteria. The proposed concept is elucidated and validated on two different data sets as examples. The first example analyzes a set of pressure data from a swirl-stabilized combustor, where controlled protocols are used to induce flame instabilities. Variations in the complexity of the derived Markov model represent how the system operating condition changes from a stable to an unstable combustion regime. In the second example, the data set is taken from NASA's data repository for prognostics of bearings on rotating shafts. We show that, even with a very small state-space, the reduced-order models are able to achieve comparable performance and that the proposed approach provides flexibility in the selection of a final model for representation and learning.

This paper proposes a minimal contrastive divergence method for learning energy-based generative ConvNet models of images at multiple grids (or scales) simultaneously. For each grid, we learn an energy-based probabilistic model where the energy function is defined by a bottom-up convolutional neural network (ConvNet or CNN). Learning such a model requires generating synthesized examples from the model. Within each iteration of our learning algorithm, for each observed training image, we generate synthesized images at multiple grids by initializing the finite-step MCMC sampling from a minimal 1 x 1 version of the training image. The synthesized image at each subsequent grid is obtained by a finite-step MCMC initialized from the synthesized image generated at the previous coarser grid. After obtaining the synthesized examples, the parameters of the models at multiple grids are updated separately and simultaneously based on the differences between synthesized and observed examples. We call this learning method the multi-grid minimal contrastive divergence. We show that this method can learn realistic energy-based generative ConvNet models, and it outperforms the original contrastive divergence (CD) and persistent CD.

Questions your company should be asking before implementing machine learning. Machine learning (ML) is all the craze right now. You hear about Elon Musk and Mark Zuckerberg debate the future of artificial intelligence and machine learning, but you wonder, how is machine learning going to actually help my business? In this article, we briefly explain what ML is and then dive into the ML-related questions your company should be asking. Machine learning is revolutionary because it gives computers the ability to solve problems without being explicitly programmed.

Learning, taking into account full distribution of the data, referred to as generative, is not feasible with deep neural networks (DNNs) because they model only the conditional distribution of the outputs given the inputs. Current solutions are either based on joint probability models facing difficult estimation problems or learn two separate networks, mapping inputs to outputs (recognition) and vice-versa (generation). We propose an intermediate approach. First, we show that forward computation in DNNs with logistic sigmoid activations corresponds to a simplified approximate Bayesian inference in a directed probabilistic multi-layer model. This connection allows to interpret DNN as a probabilistic model of the output and all hidden units given the input. Second, we propose that in order for the recognition and generation networks to be more consistent with the joint model of the data, weights of the recognition and generator network should be related by transposition. We demonstrate in a tentative experiment that such a coupled pair can be learned generatively, modelling the full distribution of the data, and has enough capacity to perform well in both recognition and generation.

Recurrent neural networks (RNNs) are a vital modeling technique that rely on internal states learned indirectly by optimization of a supervised, unsupervised, or reinforcement training loss. RNNs are used to model dynamic processes that are characterized by underlying latent states whose form is often unknown, precluding its analytic representation inside an RNN. In the Predictive-State Representation (PSR) literature, latent state processes are modeled by an internal state representation that directly models the distribution of future observations, and most recent work in this area has relied on explicitly representing and targeting sufficient statistics of this probability distribution. We seek to combine the advantages of RNNs and PSRs by augmenting existing state-of-the-art recurrent neural networks with Predictive-State Decoders (PSDs), which add supervision to the network's internal state representation to target predicting future observations. Predictive-State Decoders are simple to implement and easily incorporated into existing training pipelines via additional loss regularization. We demonstrate the effectiveness of PSDs with experimental results in three different domains: probabilistic filtering, Imitation Learning, and Reinforcement Learning. In each, our method improves statistical performance of state-of-the-art recurrent baselines and does so with fewer iterations and less data.

The edge partition model (EPM) is a fundamental Bayesian nonparametric model for extracting an overlapping structure from binary matrix. The EPM adopts a gamma process ($\Gamma$P) prior to automatically shrink the number of active atoms. However, we empirically found that the model shrinkage of the EPM does not typically work appropriately and leads to an overfitted solution. An analysis of the expectation of the EPM's intensity function suggested that the gamma priors for the EPM hyperparameters disturb the model shrinkage effect of the internal $\Gamma$P. In order to ensure that the model shrinkage effect of the EPM works in an appropriate manner, we proposed two novel generative constructions of the EPM: CEPM incorporating constrained gamma priors, and DEPM incorporating Dirichlet priors instead of the gamma priors. Furthermore, all DEPM's model parameters including the infinite atoms of the $\Gamma$P prior could be marginalized out, and thus it was possible to derive a truly infinite DEPM (IDEPM) that can be efficiently inferred using a collapsed Gibbs sampler. We experimentally confirmed that the model shrinkage of the proposed models works well and that the IDEPM indicated state-of-the-art performance in generalization ability, link prediction accuracy, mixing efficiency, and convergence speed.

Probabilistic models analyze data by relying on a set of assumptions. Data that exhibit deviations from these assumptions can undermine inference and prediction quality. Robust models offer protection against mismatch between a model's assumptions and reality. We propose a way to systematically detect and mitigate mismatch of a large class of probabilistic models. The idea is to raise the likelihood of each observation to a weight and then to infer both the latent variables and the weights from data. Inferring the weights allows a model to identify observations that match its assumptions and down-weight others. This enables robust inference and improves predictive accuracy. We study four different forms of mismatch with reality, ranging from missing latent groups to structure misspecification. A Poisson factorization analysis of the Movielens 1M dataset shows the benefits of this approach in a practical scenario.