This review deals with Restricted Boltzmann Machine (RBM) under the light of statistical physics. The RBM is a classical family of Machine learning (ML) models which played a central role in the development of deep learning. Viewing it as a Spin Glass model and exhibiting various links with other models of statistical physics, we gather recent results dealing with mean-field theory in this context. First the functioning of the RBM can be analyzed via the phase diagrams obtained for various statistical ensembles of RBM leading in particular to identify a {\it compositional phase} where a small number of features or modes are combined to form complex patterns. Then we discuss recent works either able to devise mean-field based learning algorithms; either able to reproduce generic aspects of the learning process from some {\it ensemble dynamics equations} or/and from linear stability arguments.

We propose a new method in which a generative adversarial network (GAN) is used to quantify the uncertainty of forward simulations in the presence of observed data. Previously, a method has been developed which enables GANs to make time series predictions and data assimilation by training a GAN with unconditional simulations of a high-fidelity numerical model. After training, the GAN can be used to predict the evolution of the spatial distribution of the simulation states and observed data is assimilated. In this paper, we describe the process required in order to quantify uncertainty, during which no additional simulations of the high-fidelity numerical model are required. These methods take advantage of the adjoint-like capabilities of generative models and the ability to simulate forwards and backwards in time. Set within a reduced-order model framework for efficiency, we apply these methods to a compartmental model in epidemiology to predict the spread of COVID-19 in an idealised town. The results show that the proposed method can efficiently quantify uncertainty in the presence of measurements using only unconditional simulations of the high-fidelity numerical model.

A well-trained DNN model can be regarded as an intellectual property (IP) of the model owner. To date, many DNN IP protection methods have been proposed, but most of them are watermarking based verification methods where model owners can only verify their ownership passively after the copyright of DNN models has been infringed. In this paper, we propose an effective framework to actively protect the DNN IP from infringement. Specifically, we encrypt the DNN model's parameters by perturbing them with well-crafted adversarial perturbations. With the encrypted parameters, the accuracy of the DNN model drops significantly, which can prevent malicious infringers from using the model. After the encryption, the positions of encrypted parameters and the values of the added adversarial perturbations form a secret key. Authorized user can use the secret key to decrypt the model. Compared with the watermarking methods which only passively verify the ownership after the infringement occurs, the proposed method can prevent infringement in advance. Moreover, compared with most of the existing active DNN IP protection methods, the proposed method does not require additional training process of the model, which introduces low computational overhead. Experimental results show that, after the encryption, the test accuracy of the model drops by 80.65%, 81.16%, and 87.91% on Fashion-MNIST, CIFAR-10, and GTSRB, respectively. Moreover, the proposed method only needs to encrypt an extremely low number of parameters, and the proportion of the encrypted parameters of all the model's parameters is as low as 0.000205%. The experimental results also indicate that, the proposed method is robust against model fine-tuning attack and model pruning attack. Moreover, for the adaptive attack where attackers know the detailed steps of the proposed method, the proposed method is also demonstrated to be robust.

We train deep generative models on datasets of reflexive polytopes. This enables us to compare how well the models have picked up on various global properties of generated samples. Our datasets are complete in the sense that every single example, up to changes of coordinate, is included in the dataset. Using this property we also perform tests checking to what extent the models are merely memorizing the data. We also train models on the same dataset represented in two different ways, enabling us to measure which form is easiest to learn from. We use these experiments to show that deep generative models can learn to generate geometric objects with non-trivial global properties, and that the models learn some underlying properties of the objects rather than simply memorizing the data.

This paper addresses the energy management of a grid-connected renewable generation plant coupled with a battery energy storage device in the capacity firming market, designed to promote renewable power generation facilities in small non-interconnected grids. A recently developed deep learning model known as normalizing flows is used to generate quantile forecasts of renewable generation. They provide a general mechanism for defining expressive probability distributions, only requiring the specification of a base distribution and a series of bijective transformations. Then, a probabilistic forecast-driven strategy is designed, modeled as a min-max-min robust optimization problem with recourse, and solved using a Benders decomposition. The convergence is improved by building an initial set of cuts derived from domain knowledge. Robust optimization models the generation randomness using an uncertainty set that includes the worst-case generation scenario and protects this scenario under the minimal increment of costs. This approach improves the results over a deterministic approach with nominal point forecasts by finding a trade-off between conservative and risk-seeking policies. Finally, a dynamic risk-averse parameters selection strategy based on the quantile forecasts distribution provides an additional gain. The case study uses the photovoltaic generation monitored on-site at the University of Li\`ege (ULi\`ege), Belgium.

EEG is the gold standard for seizure detection in the newborn infant, but EEG interpretation in the preterm group is particularly challenging; trained experts are scarce and the task of interpreting EEG in real-time is arduous. Preterm infants are reported to have a higher incidence of seizures compared to term infants. Preterm EEG morphology differs from that of term infants, which implies that seizure detection algorithms trained on term EEG may not be appropriate. The task of developing preterm specific algorithms becomes extra-challenging given the limited amount of annotated preterm EEG data available. This paper explores novel deep learning (DL) architectures for the task of neonatal seizure detection in preterm infants. The study tests and compares several approaches to address the problem: training on data from full-term infants; training on data from preterm infants; training on age-specific preterm data and transfer learning. The system performance is assessed on a large database of continuous EEG recordings of 575h in duration. It is shown that the accuracy of a validated term-trained EEG seizure detection algorithm, based on a support vector machine classifier, when tested on preterm infants falls well short of the performance achieved for full-term infants. An AUC of 88.3% was obtained when tested on preterm EEG as compared to 96.6% obtained when tested on term EEG. When re-trained on preterm EEG, the performance marginally increases to 89.7%. An alternative DL approach shows a more stable trend when tested on the preterm cohort, starting with an AUC of 93.3% for the term-trained algorithm and reaching 95.0% by transfer learning from the term model using available preterm data.

Analysis of large observational data sets generated by a reactive system is a common challenge in debugging system failures and determining their root cause. One of the major problems is that these observational data suffer from survivorship bias. Examples include analyzing traffic logs from networks, and simulation logs from circuit design. In such applications, users want to detect non-spurious correlations from observational data and obtain actionable insights about them. In this paper, we introduce log to Neuro-symbolic (Log2NS), a framework that combines probabilistic analysis from machine learning (ML) techniques on observational data with certainties derived from symbolic reasoning on an underlying formal model. We apply the proposed framework to network traffic debugging by employing the following steps. To detect patterns in network logs, we first generate global embedding vector representations of entities such as IP addresses, ports, and applications. Next, we represent large log flow entries as clusters that make it easier for the user to visualize and detect interesting scenarios that will be further analyzed. To generalize these patterns, Log2NS provides an ability to query from static logs and correlation engines for positive instances, as well as formal reasoning for negative and unseen instances. By combining the strengths of deep learning and symbolic methods, Log2NS provides a very powerful reasoning and debugging tool for log-based data. Empirical evaluations on a real internal data set demonstrate the capabilities of Log2NS.

Reinforcement learning agents often forget details of the past, especially after delays or distractor tasks. Agents with common memory architectures struggle to recall and integrate across multiple timesteps of a past event, or even to recall the details of a single timestep that is followed by distractor tasks. To address these limitations, we propose a Hierarchical Transformer Memory (HTM), which helps agents to remember the past in detail. HTM stores memories by dividing the past into chunks, and recalls by first performing high-level attention over coarse summaries of the chunks, and then performing detailed attention within only the most relevant chunks. An agent with HTM can therefore "mentally time-travel" -- remember past events in detail without attending to all intervening events. We show that agents with HTM substantially outperform agents with other memory architectures at tasks requiring long-term recall, retention, or reasoning over memory. These include recalling where an object is hidden in a 3D environment, rapidly learning to navigate efficiently in a new neighborhood, and rapidly learning and retaining new object names. Agents with HTM can extrapolate to task sequences an order of magnitude longer than they were trained on, and can even generalize zero-shot from a meta-learning setting to maintaining knowledge across episodes. HTM improves agent sample efficiency, generalization, and generality (by solving tasks that previously required specialized architectures). Our work is a step towards agents that can learn, interact, and adapt in complex and temporally-extended environments.

Social network platforms are generally used to share positive, constructive, and insightful content. However, in recent times, people often get exposed to objectionable content like threat, identity attacks, hate speech, insults, obscene texts, offensive remarks or bullying. Existing work on toxic speech detection focuses on binary classification or on differentiating toxic speech among a small set of categories. This paper describes the system proposed by team Cisco for SemEval-2021 Task 5: Toxic Spans Detection, the first shared task focusing on detecting the spans in the text that attribute to its toxicity, in English language. We approach this problem primarily in two ways: a sequence tagging approach and a dependency parsing approach. In our sequence tagging approach we tag each token in a sentence under a particular tagging scheme. Our best performing architecture in this approach also proved to be our best performing architecture overall with an F1 score of 0.6922, thereby placing us 7th on the final evaluation phase leaderboard. We also explore a dependency parsing approach where we extract spans from the input sentence under the supervision of target span boundaries and rank our spans using a biaffine model. Finally, we also provide a detailed analysis of our results and model performance in our paper.

We are using NVIDIA Jetson Nano Developer kit as our accelerator system.Which will contain Docker Container which will contain the dataset and trained model SSD (Single Shot Detector) MobileNetV2 which we will be used to for facial detection. Video would be recorded through the Camera attached to the accelerator system. Code of the SSD (Single Shot Detector) MobileNetV2 is written in Python Programming Language and Deep learning framework which has been used is PyTorch.To optimized the neural network layers.NVIDIA TensorRT is used for faster Inference during the run time.