SMILES Transformer: Pre-trained Molecular Fingerprint for Low Data Drug Discovery Shion Honda 1,2,3, Shoi Shi 1,2,3, Hiroki R. Ueda 1,2,3 1 University of Tokyo 2 International Research Center for Neurointelligence 3 RIKEN Center for Biosystems Dynamics Research shion honda@ipc.i.u-tokyo.ac.jp, { sshoi0322-tky,uedah-tky}@umin.ac.jp Abstract In drug-discovery-related tasks such as virtual screening, machine learning is emerging as a promising way to predict molecular properties. Conventionally, molecular fingerprints (numerical representations of molecules) are calculated through rule-based algorithms that map molecules to a sparse discrete space. However, these algorithms perform poorly for shallow prediction models or small datasets. To address this issue, we present SMILES Transformer. Inspired by Transformer and pre-trained language models from natural language processing, SMILES Transformer learns molecular fingerprints through unsupervised pre-training of the sequence-to-sequence language model using a huge corpus of SMILES, a text representation system for molecules. We performed benchmarks on 10 datasets against existing fingerprints and graph-based methods and demonstrated the superiority of the proposed algorithms in small-data settings where pre-training facilitated good generalization.

Integrating ML models in software is of growing interest. Building accurate models requires right choice of hyperparameters for training procedures (learners), when the training dataset is given. AutoML tools provide APIs to automate the choice, which usually involve many trials of different hyperparameters for a given training dataset. Since training and evaluation of complex models can be time and resource consuming, existing AutoML solutions require long time or large resource to produce accurate models for large scale training data. That prevents AutoML to be embedded in a software which needs to repeatedly tune hyperparameters and produce models to be consumed by other components, such as large-scale data systems. We present a fast and lightweight hyperparameter optimization method FLO and use it to build an efficient AutoML solution. Our method optimizes for minimal evaluation cost instead of number of iterations to find accurate models. Our main idea is to leverage a holistic consideration of the relations among model complexity, evaluation cost and accuracy. FLO has a strong anytime performance and significantly outperforms Bayesian Optimization and random search for hyperparameter tuning on a large open source AutoML Benchmark. Our AutoML solution also outperforms top-ranked AutoML libraries in a majority of the tasks on this benchmark.

Text clustering is arguably one of the most important topics in modern data mining. Nevertheless, text data require tokenization which usually yields a very large and highly sparse term-document matrix, which is usually difficult to process using conventional machine learning algorithms. Methods such as Latent Semantic Analysis have helped mitigate this issue, but are nevertheless not completely stable in practice. As a result, we propose a new feature agglomeration method based on Nonnegative Matrix Factorization. NMF is employed to separate the terms into groups, and then each group`s term vectors are agglomerated into a new feature vector. Together, these feature vectors create a new feature space much more suitable for clustering. In addition, we propose a new deterministic initialization for spherical K-Means, which proves very useful for this specific type of data. In order to evaluate the proposed method, we compare it to some of the latest research done in this field, as well as some of the most practiced methods. In our experiments, we conclude that the proposed method either significantly improves clustering performance, or maintains the performance of other methods, while improving stability in results.

Advances in deep generative and density models have shown impressive capacity to model complex probability density functions in lower-dimensional space. Also, applying such models to high-dimensional image data to model the PDF has shown poor generalization, with out-of-distribution data being assigned equal or higher likelihood than in-sample data. Methods to deal with this have been proposed that deviate from a fully unsupervised approach, requiring large ensembles or additional knowledge about the data, not commonly available in the real-world. In this work, the previously offered reasoning behind these issues is challenged empirically, and it is shown that data-sets such as MNIST fashion/digits and CIFAR10/SVHN are trivially separable and have no overlap on their respective data manifolds that explains the higher OoD likelihood. Models like masked autoregressive flows and block neural autoregressive flows are shown to not suffer from OoD likelihood issues to the extent of GLOW, PixelCNN++, and real NVP. A new avenue is also explored which involves a change of basis to a new space of the same dimension with an orthonormal unitary basis of eigenvectors before modeling. In the test data-sets and models, this aids in pushing down the relative likelihood of the contrastive OoD data set and improve discrimination results. The significance of the density of the original space is maintained, while invertibility remains tractable. Finally, a look to the previous generation of generative models in the form of probabilistic principal component analysis is inspired, and revisited for the same data-sets and shown to work really well for discriminating anomalies based on likelihood in a fully unsupervised fashion compared with pixelCNN++, GLOW, and real NVP with less complexity and faster training. Also, dimensionality reduction using PCA is shown to improve anomaly detection in generative models.

Meta-learning for few-shot learning allows a machine to leverage previously acquired knowledge as a prior, thus improving the performance on novel tasks with only small amounts of data. However, most mainstream models suffer from catastrophic forgetting and insufficient robustness issues, thereby failing to fully retain or exploit long-term knowledge while being prone to cause severe error accumulation. In this paper, we propose a novel Continual Meta-Learning approach with Bayesian Graph Neural Networks (CML-BGNN) that mathematically formulates meta-learning as continual learning of a sequence of tasks. With each task forming as a graph, the intra- and inter-task correlations can be well preserved via message-passing and history transition. To remedy topological uncertainty from graph initialization, we utilize Bayes by Backprop strategy that approximates the posterior distribution of task-specific parameters with amortized inference networks, which are seamlessly integrated into the end-to-end edge learning. Extensive experiments conducted on the miniImageNet and tieredImageNet datasets demonstrate the effectiveness and efficiency of the proposed method, improving the performance by 42.8% compared with state-of-the-art on the miniImageNet 5-way 1-shot classification task.

We study the design of computationally efficient algorithms with provable guarantees, that are robust to adversarial (test time) perturbations. While there has been an proliferation of recent work on this topic due to its connections to test time robustness of deep networks, there is limited theoretical understanding of several basic questions like (i) when and how can one design provably robust learning algorithms? (ii) what is the price of achieving robustness to adversarial examples in a computationally efficient manner? The main contribution of this work is to exhibit a strong connection between achieving robustness to adversarial examples, and a rich class of polynomial optimization problems, thereby making progress on the above questions. In particular, we leverage this connection to (a) design computationally efficient robust algorithms with provable guarantees for a large class of hypothesis, namely linear classifiers and degree-2 polynomial threshold functions (PTFs), (b) give a precise characterization of the price of achieving robustness in a computationally efficient manner for these classes, (c) design efficient algorithms to certify robustness and generate adversarial attacks in a principled manner for 2-layer neural networks. We empirically demonstrate the effectiveness of these attacks on real data.

Early detection is a crucial goal in the study of Alzheimer's Disease (AD). In this work, we describe several techniques to boost the performance of 3D convolutional neural networks trained to detect AD using structural brain MRI scans. Specifically, we provide evidence that (1) instance normalization outperforms batch normalization, (2) early spatial downsampling negatively affects performance, (3) widening the model brings consistent gains while increasing the depth does not, and (4) incorporating age information yields moderate improvement. Together, these insights yield an increment of approximately 14% in test accuracy over existing models when distinguishing between patients with AD, mild cognitive impairment, and controls in the ADNI dataset. Similar performance is achieved on an independent dataset.

Conditional Random Fields (CRF) are frequently applied for labeling and segmenting sequence data. Morency et al. (2007) introduced hidden state variables in a labeled CRF structure in order to model the latent dynamics within class labels, thus improving the labeling performance. Such a model is known as Latent-Dynamic CRF (LDCRF). We present Factored LDCRF (FLDCRF), a structure that allows multiple latent dynamics of the class labels to interact with each other. Including such latent-dynamic interactions leads to improved labeling performance on single-label and multi-label sequence modeling tasks. We apply our FLDCRF models on two single-label (one nested cross-validation) and one multi-label sequence tagging (nested cross-validation) experiments across two different datasets - UCI gesture phase data and UCI opportunity data. FLDCRF outperforms all state-of-the-art sequence models, i.e., CRF, LDCRF, LSTM, LSTM-CRF, Factorial CRF, Coupled CRF and a multi-label LSTM model in all our experiments. In addition, LSTM based models display inconsistent performance across validation and test data, and pose diffculty to select models on validation data during our experiments. FLDCRF offers easier model selection, consistency across validation and test performance and lucid model intuition. FLDCRF is also much faster to train compared to LSTM, even without a GPU. FLDCRF outshines the best LSTM model by ~4% on a single-label task on UCI gesture phase data and outperforms LSTM performance by ~2% on average across nested cross-validation test sets on the multi-label sequence tagging experiment on UCI opportunity data. The idea of FLDCRF can be extended to joint (multi-agent interactions) and heterogeneous (discrete and continuous) state space models.

Sleep stage classification constitutes an important element of sleep disorder diagnosis. It relies on the visual inspection of polysomnography records by trained sleep technologists. Automated approaches have been designed to alleviate this resource-intensive task. However, such approaches are usually compared to a single human scorer annotation despite an inter-rater agreement of about 85 % only. The present study introduces two publicly-available datasets, DOD-H including 25 healthy volunteers and DOD-O including 55 patients suffering from obstructive sleep apnea (OSA). Both datasets have been scored by 5 sleep technologists from different sleep centers. We developed a framework to compare automated approaches to a consensus of multiple human scorers. Using this framework, we benchmarked and compared the main literature approaches. We also developed and benchmarked a new deep learning method, SimpleSleepNet, inspired by current state-of-the-art. We demonstrated that many methods can reach human-level performance on both datasets. SimpleSleepNet achieved an F1 of 89.9 % vs 86.8 % on average for human scorers on DOD-H, and an F1 of 88.3 % vs 84.8 % on DOD-O. Our study highlights that using state-of-the-art automated sleep staging outperforms human scorers performance for healthy volunteers and patients suffering from OSA. Consideration could be made to use automated approaches in the clinical setting.

Pathology reports contain useful information such as the main involved organ, diagnosis, etc. These information can be identified from the free text reports and used for large-scale statistical analysis or serve as annotation for other modalities such as pathology slides images. However, manual classification for a huge number of reports on multiple tasks is labor-intensive. In this paper, we have developed an automatic text classifier based on BERT and we propose a human-centric metric to evaluate the model. According to the model confidence, we identify low-confidence cases that require further expert annotation and high-confidence cases that are automatically classified. We report the percentage of low-confidence cases and the performance of automatically classified cases. On the high-confidence cases, the model achieves classification accuracy comparable to pathologists. This leads a potential of reducing 80% to 98% of the manual annotation workload.