We present an accurate, fast and efficient method for segmentation and muscle mask propagation in 3D freehand ultrasound data, towards accurate volume quantification. To this end, we propose a deep Siamese 3D Encoder-Decoder network that captures the evolution of the muscle appearance and shape for contiguous slices and uses it to propagate a reference mask annotated by a clinical expert. To handle longer changes of the muscle shape over the entire volume and to provide an accurate propagation, we devised a Bidirectional Long Short Term Memory module. To train our model with a minimal amount of training samples, we propose a strategy to combine learning from few annotated 2D ultrasound slices with sequential pseudo-labeling of the unannotated slices. To promote few-shot learning, we propose a decremental update of the objective function to guide the model convergence in the absence of large amounts of annotated data. Finally, to handle the class-imbalance between foreground and background muscle pixels, we propose a parametric Tversky loss function that learns to adaptively penalize false positives and false negatives. We validate our approach for the segmentation, label propagation, and volume computation of the three low-limb muscles on a dataset of 44 subjects. We achieve a dice score coefficient of over $95~\%$ and a small fraction of error with $1.6035~\pm~0.587$.

We apply a Transformer architecture, specifically BERT, to learn flexible and high quality molecular representations for drug discovery problems. We study the impact of using different combinations of self-supervised tasks for pre-training, and present our results for the established Virtual Screening and QSAR benchmarks. We show that: i) The selection of appropriate self-supervised task(s) for pre-training has a significant impact on performance in subsequent downstream tasks such as Virtual Screening. ii) Using auxiliary tasks with more domain relevance for Chemistry, such as learning to predict calculated molecular properties, increases the fidelity of our learnt representations. iii) Finally, we show that molecular representations learnt by our model `MolBert' improve upon the current state of the art on the benchmark datasets.

Remarkable success of modern image-based AI methods and the resulting interest in their applications in critical decision-making processes has led to a surge in efforts to make such intelligent systems transparent and explainable. The need for explainable AI does not stem only from ethical and moral grounds but also from stricter legislation around the world mandating clear and justifiable explanations of any decision taken or assisted by AI. Especially in the medical context where Computer-Aided Diagnosis can have a direct influence on the treatment and well-being of patients, transparency is of utmost importance for safe transition from lab research to real world clinical practice. This paper provides a comprehensive overview of current state-of-the-art in explaining and interpreting Deep Learning based algorithms in applications of medical research and diagnosis of diseases. We discuss early achievements in development of explainable AI for validation of known disease criteria, exploration of new potential biomarkers, as well as methods for the subsequent correction of AI models. Various explanation methods like visual, textual, post-hoc, ante-hoc, local and global have been thoroughly and critically analyzed. Subsequently, we also highlight some of the remaining challenges that stand in the way of practical applications of AI as a clinical decision support tool and provide recommendations for the direction of future research.

Our contributions (NLP) with less supervision or without supervision are as follows to illustrate our explorations in has been a continuously attractive problem. The unsupervised how to improve clustering accuracy: approaches can be roughly classified as - We find that to additionally introduce generative and discriminative (Chen et al., 2020). A UDA (Xie et al., 2019) in CL can further improve generative approach may use an encoding network clustering accuracy; to learn latent representations for input texts, and feeds the latent representations into another generating - Our experimental results suggest that PCL network. Given an objective of generation can achieve almost the same performance similarity, the most proper latent representations with supervised learning using partial items in can be learned. The generative approaches always dataset; introduce novel neural architectures, for example - We also find that using multi-language back Chiu et al. (Chiu et al., 2020) compared a series translation to generate positive sample in SCL of graph neural network in document clustering can further eliminate the contrast processing accuraccy of 20newsgroup (Lang, 1995). On the in PCL with some performance deteriorations; contrary, the models that are commonly seen in supervised learning are preferred to be reused for a discriminative method.

Artificial intelligence for autonomous driving must meet strict requirements on safety and robustness. We propose to validate machine learning models for self-driving vehicles not only with given ground truth labels, but also with additional a-priori knowledge. In particular, we suggest to validate the drivable area in semantic segmentation masks using given street map data. We present first results, which indicate that prediction errors can be uncovered by map-based validation.

Machine learning (ML) models have been successfully applied to the analysis of sensitive user data such as medical images (Lundervold & Lundervold, 2019), text messages (Chen et al., 2019) or social media posts (Wu et al., 2016). Training these ML models under the framework of differential privacy (DP) (Dwork et al., 2006b; Chaudhuri et al., 2011; Shokri & Shmatikov, 2015; Abadi et al., 2016) can protect deployed classifiers against unintentional leakage of private training data (Shokri et al., 2017; Song et al., 2017; Carlini et al., 2019). Yet, training deep neural networks with strong DP guarantees comes at a significant cost in utility (Abadi et al., 2016; Yu et al., 2020; Bagdasaryan et al., 2019; Feldman, 2020). In fact, on many ML benchmarks the reported accuracy of private deep learning still falls short of "shallow" (non-private) techniques. For example, on CIFAR-10, Papernot et al. (2020b) train a neural network to 66.2% accuracy for a large DP budget of ε 7.53, the highest accuracy we are aware of for this privacy budget. Yet, without privacy, higher accuracy is achievable with linear models and non-learned "handcrafted" features, e.g., (Coates & Ng, 2012; Oyallon & Mallat, 2015). This leads to the central question of our work: Can differentially private learning benefit from handcrafted features? We answer this question affirmatively by introducing simple and strong handcrafted baselines for differentially private learning, that significantly improve the privacy-utility guarantees on canonical vision benchmarks.

In this work, we propose CBiGAN -- a novel method for anomaly detection in images, where a consistency constraint is introduced as a regularization term in both the encoder and decoder of a BiGAN. Our model exhibits fairly good modeling power and reconstruction consistency capability. We evaluate the proposed method on MVTec AD -- a real-world benchmark for unsupervised anomaly detection on high-resolution images -- and compare against standard baselines and state-of-the-art approaches. Experiments show that the proposed method improves the performance of BiGAN formulations by a large margin and performs comparably to expensive state-of-the-art iterative methods while reducing the computational cost. We also observe that our model is particularly effective in texture-type anomaly detection, as it sets a new state of the art in this category. Our code is available at https://github.com/fabiocarrara/cbigan-ad/.

The most frightening potential use is disinformation. Imagine a computer that can create unlimited amounts of false information or fake photos and you can't distinguish between what's real and fabricated. We hear all the time about technology that writes like people or converses like us, or makes realistic computer-generated human faces or faked videos. Repeatedly since the 1950s scientists have produced technologies that seemed as if we were close to mimicking human intelligence. Each time we were nowhere close.

It's (mostly) true that Data Scientists do not care about infrastructure. Indeed, even though DevOps is a very interesting field, most of them are not exactly eager to start a VM, allocate the needed resources, configure the network, ssh into the machine, build a docker image and launch a Jupyter Notebook server. To cut to the chase, in this story, we create a ready to use, GPU accelerated Deep Learning environment, that has already TensorFlow and PyTorch installed. To do that we need to create the Dockerfile that describes the environment, build it and use it as the image of the Notebook server inside a Kubeflow instance. So, without further ado let's see the Dockerfile and walk through it step by step.

Our dataset is imbalanced (5,000 masked faces VS 90,000 non-masked faces). Therefore, when splitting the dataset into train/validation, we need to keep the same proportions of the samples in train/validation as the whole dataset. We do that by using the train_test_split function of sklearn and we pass the dataset's labels to its stratisfy parameter, and it will do the rest for us. We're going to use 70% of the dataset for training and 30% for validation: When dealing with unbalanced data, we need to pass this information to the loss function to avoid unproportioned step sizes of the optimizer. We do this by assigning a weight to each class, according to its representability in the dataset.