In the regression problem, we consider the problem of estimating the variance function by the means of aggregation methods. We focus on two particular aggregation setting: Model Selection aggregation (MS) and Convex aggregation (C) where the goal is to select the best candidate and to build the best convex combination of candidates respectively among a collection of candidates. In both cases, the construction of the estimator relies on a two-step procedure and requires two independent samples. The first step exploits the first sample to build the candidate estimators for the variance function by the residual-based method and then the second dataset is used to perform the aggregation step. We show the consistency of the proposed method with respect to the L 2error both for MS and C aggregations. We evaluate the performance of these two methods in the heteroscedastic model and illustrate their interest in the regression problem with reject option.

Graph neural networks (GNNs) have been shown to possess strong representation power, which can be exploited for downstream prediction tasks on graph-structured data, such as molecules and social networks. They typically learn representations by aggregating information from the K-hop neighborhood of individual vertices or from the enumerated walks in the graph. Prior studies have demonstrated the effectiveness of incorporating weighting schemes into GNNs; however, this has been primarily limited to K-hop neighborhood GNNs so far. In this paper, we aim to extensively analyze the effect of incorporating weighting schemes into walk-aggregating GNNs. Towards this objective, we propose a novel GNN model, called AWARE, that aggregates information about the walks in the graph using attention schemes in a principled way to obtain an end-to-end supervised learning method for graph-level prediction tasks. We perform theoretical, empirical, and interpretability analyses of AWARE. Our theoretical analysis provides the first provable guarantees for weighted GNNs, demonstrating how the graph information is encoded in the representation, and how the weighting schemes in AWARE affect the representation and learning performance. We empirically demonstrate the superiority of AWARE over prior baselines in the domains of molecular property prediction (61 tasks) and social networks (4 tasks). Our interpretation study illustrates that AWARE can successfully learn to capture the important substructures of the input graph.

Graph is an important data representation ubiquitously existing in the real world. However, analyzing the graph data is computationally difficult due to its non-Euclidean nature. Graph embedding is a powerful tool to solve the graph analytics problem by transforming the graph data into low-dimensional vectors. These vectors could also be shared with third parties to gain additional insights of what is behind the data. While sharing graph embedding is intriguing, the associated privacy risks are unexplored. In this paper, we systematically investigate the information leakage of the graph embedding by mounting three inference attacks. First, we can successfully infer basic graph properties, such as the number of nodes, the number of edges, and graph density, of the target graph with up to 0.89 accuracy. Second, given a subgraph of interest and the graph embedding, we can determine with high confidence that whether the subgraph is contained in the target graph. For instance, we achieve 0.98 attack AUC on the DD dataset. Third, we propose a novel graph reconstruction attack that can reconstruct a graph that has similar graph structural statistics to the target graph. We further propose an effective defense mechanism based on graph embedding perturbation to mitigate the inference attacks without noticeable performance degradation for graph classification tasks. Our code is available at https://github.com/Zhangzhk0819/GNN-Embedding-Leaks.

Uncertainty estimation in deep learning has recently emerged as a crucial area of interest to advance reliability and robustness in safety-critical applications. While there have been many proposed methods that either focus on distance-aware model uncertainties for out-of-distribution detection or on input-dependent label uncertainties for in-distribution calibration, both of these types of uncertainty are often necessary. In this work, we propose the HetSNGP method for jointly modeling the model and data uncertainty. We show that our proposed model affords a favorable combination between these two complementary types of uncertainty and thus outperforms the baseline methods on some challenging out-of-distribution datasets, including CIFAR-100C, Imagenet-C, and Imagenet-A. Moreover, we propose HetSNGP Ensemble, an ensembled version of our method which adds an additional type of uncertainty and also outperforms other ensemble baselines.

The rapid growth in the amount of data processed by analysts demands more efficient information digestion and communication methods. Data visualization by dimensionality reduction facilitates a viewer to digest information in massive data sets quickly. Therefore, it is increasingly applied as a critical component in scientific research, digital libraries, data mining, financial data analysis, market studies, manufacturing production control, drug discovery, etc. Stochastic Neighbor Embedding (SNE) [4] is a widely used nonlinear dimensionality reduction (NLDR) method, which approximately preserves the pairwise probabilities of being neighbors (neighboring probabilities for short) in the input space. In particular, the Student t-Distributed Stochastic Neighbor Embedding (t-SNE) [9] has become one of the most popular nonlinear dimensionality reduction methods for data visualization. The t-SNE method employs a heavy-tailed distribution for the neighboring probabilities in the embedding and minimizes their Kullback-Leibler divergence against the precomputed input probabilities.

The training of sparse neural networks is becoming an increasingly important tool for reducing the computational footprint of models at training and evaluation, as well enabling the effective scaling up of models. Whereas much work over the years has been dedicated to specialised pruning techniques, little attention has been paid to the inherent effect of gradient based training on model sparsity. In this work, we introduce Powerpropagation, a new weight-parameterisation for neural networks that leads to inherently sparse models. Exploiting the behaviour of gradient descent, our method gives rise to weight updates exhibiting a "rich get richer" dynamic, leaving low-magnitude parameters largely unaffected by learning. Models trained in this manner exhibit similar performance, but have a distribution with markedly higher density at zero, allowing more parameters to be pruned safely. Powerpropagation is general, intuitive, cheap and straight-forward to implement and can readily be combined with various other techniques. To highlight its versatility, we explore it in two very different settings: Firstly, following a recent line of work, we investigate its effect on sparse training for resource-constrained settings. Here, we combine Powerpropagation with a traditional weight-pruning technique as well as recent state-of-the-art sparse-to-sparse algorithms, showing superior performance on the ImageNet benchmark. Secondly, we advocate the use of sparsity in overcoming catastrophic forgetting, where compressed representations allow accommodating a large number of tasks at fixed model capacity. In all cases our reparameterisation considerably increases the efficacy of the off-the-shelf methods.

Machine learning (ML) is the process of training a computer-related program that helps to create a statistical model based on data. It automatically learns programs from data. Machine learning is one of the by-products of artificial intelligence (AI). Nowadays, almost 80% of enterprises already adopt machine learning and artificial intelligence and have gained enormous financial advantages from it. So, let us quickly look into these top 20 interview questions with answers which may help you to crack your interview.

Acquiring the required data science skills has become a very prominent essential if you're looking to become a data scientist. As such, data scientist is one of the top-notch job roles wherein highly essential skilled aspirants are queuing to work! While many of you may have the right abilities and expertise, it is also important that you showcase them well. Data Science is a highly complex and niche segment and it requires a bag full of skills. Amongst them there are a few qualities that are must have on your resumes.

The AI and Machine Learning industry is booming like never before. As of 2021, the increase in AI usage across businesses will create $2.9 trillion of business value. AI has automated many industries across the globe and changed the way they operate. Most large companies incorporate AI to maximize productivity in their workflow, and industries like marketing and healthcare have undergone a paradigm shift due to the consolidation of AI. Due to this, there has been an increasing demand in the past few years for AI professionals. There has almost been a 100% increase in AI and machine learning-related job postings from 2015 to 2018. This number has grown since and is projected to rise in 2021. If you are looking to break into the machine learning industry, the good news is that there is no shortage of jobs available. Companies need a talented workforce that is capable of pioneering the shift to machine learning.

In this paper, we argue that the unsatisfactory out-of-distribution (OOD) detection performance of neural networks is mainly due to the SoftMax loss anisotropy and propensity to produce low entropy probability distributions in disagreement with the principle of maximum entropy. Current out-of-distribution (OOD) detection approaches usually do not directly fix the SoftMax loss drawbacks, but rather build techniques to circumvent it. Unfortunately, those methods usually produce undesired side effects (e.g., classification accuracy drop, additional hyperparameters, slower inferences, and collecting extra data). In the opposite direction, we propose replacing SoftMax loss with a novel loss function that does not suffer from the mentioned weaknesses. The proposed IsoMax loss is isotropic (exclusively distance-based) and provides high entropy posterior probability distributions. Replacing the SoftMax loss by IsoMax loss requires no model or training changes. Additionally, the models trained with IsoMax loss produce as fast and energy-efficient inferences as those trained using SoftMax loss. Moreover, no classification accuracy drop is observed. The proposed method does not rely on outlier/background data, hyperparameter tuning, temperature calibration, feature extraction, metric learning, adversarial training, ensemble procedures, or generative models. Our experiments showed that IsoMax loss works as a seamless SoftMax loss drop-in replacement that significantly improves neural networks' OOD detection performance. Hence, it may be used as a baseline OOD detection approach to be combined with current or future OOD detection techniques to achieve even higher results.