The Effects of Speed and Delays on Test-Time Performance of End-to-End Self-Driving.

This study investigates the effects of speed variations and computational delays on the performance of end-to-end autonomous driving systems (ADS). Utilizing 1:10 scale mini-cars with limited computational resources, we demonstrate that different driving speeds significantly alter the task of the driving model, challenging the generalization capabilities of systems trained at a singular speed profile. Our findings reveal that models trained to drive at high speeds struggle with slower speeds and vice versa. Consequently, testing an ADS at an inappropriate speed can lead to misjudgments about its competence. Additionally, we explore the impact of computational delays, common in real-world deployments, on driving performance. We present a novel approach to counteract the effects of delays by adjusting the target labels in the training data, demonstrating improved resilience in models to handle computational delays effectively. This method, crucially, addresses the effects of delays rather than their causes and complements traditional delay minimization strategies. These insights are valuable for developing robust autonomous driving systems capable of adapting to varying speeds and delays in real-world scenarios.

LiDAR-as-Camera for End-to-End Driving.

The core task of any autonomous driving system is to transform sensory inputs into driving commands. In end-to-end driving, this is achieved via a neural network, with one or multiple cameras as the most commonly used input and low-level driving commands, e.g., steering angle, as output. However, simulation studies have shown that depth-sensing can make the end-to-end driving task easier. On a real car, combining depth and visual information can be challenging due to the difficulty of obtaining good spatial and temporal alignment of the sensors. To alleviate alignment problems, Ouster LiDARs can output surround-view LiDAR images with depth, intensity, and ambient radiation channels. These measurements originate from the same sensor, rendering them perfectly aligned in time and space. The main goal of our study is to investigate how useful such images are as inputs to a self-driving neural network. We demonstrate that such LiDAR images are sufficient for the real-car road-following task. Models using these images as input perform at least as well as camera-based models in the tested conditions. Moreover, LiDAR images are less sensitive to weather conditions and lead to better generalization. In a secondary research direction, we reveal that the temporal smoothness of off-policy prediction sequences correlates with the actual on-policy driving ability equally well as the commonly used mean absolute error.

Efficient neural decoding of self-location with a deep recurrent network.

Place cells in the mammalian hippocampus signal self-location with sparse spatially stable firing fields. Based on observation of place cell activity it is possible to accurately decode an animal's location. The precision of this decoding sets a lower bound for the amount of information that the hippocampal population conveys about the location of the animal. In this work we use a novel recurrent neural network (RNN) decoder to infer the location of freely moving rats from single unit hippocampal recordings. RNNs are biologically plausible models of neural circuits that learn to incorporate relevant temporal context without the need to make complicated assumptions about the use of prior information to predict the current state. When decoding animal position from spike counts in 1D and 2D-environments, we show that the RNN consistently outperforms a standard Bayesian approach with either flat priors or with memory. In addition, we also conducted a set of sensitivity analysis on the RNN decoder to determine which neurons and sections of firing fields were the most influential. We found that the application of RNNs to neural data allowed flexible integration of temporal context, yielding improved accuracy relative to the more commonly used Bayesian approaches and opens new avenues for exploration of the neural code.

Machine Learning for detection of viral sequences in human metagenomic datasets.

BACKGROUND: Detection of highly divergent or yet unknown viruses from metagenomics sequencing datasets is a major bioinformatics challenge. When human samples are sequenced, a large proportion of assembled contigs are classified as "unknown", as conventional methods find no similarity to known sequences. We wished to explore whether machine learning algorithms using Relative Synonymous Codon Usage frequency (RSCU) could improve the detection of viral sequences in metagenomic sequencing data. RESULTS: We trained Random Forest and Artificial Neural Network using metagenomic sequences taxonomically classified into virus and non-virus classes. The algorithms achieved accuracies well beyond chance level, with area under ROC curve 0.79. Two codons (TCG and CGC) were found to have a particularly strong discriminative capacity. CONCLUSION: RSCU-based machine learning techniques applied to metagenomic sequencing data can help identify a large number of putative viral sequences and provide an addition to conventional methods for taxonomic classification.

Common clinical blood and urine biomarkers for ischemic stroke: an Estonian Electronic Health Records database study.

BACKGROUND: Ischemic stroke (IS) is a major health risk without generally usable effective measures of primary prevention. Early warning signals that are easy to detect and widely available can save lives. Estonia has one nation-wide Electronic Health Record (EHR) database for the storage of medical information of patients from hospitals and primary care providers. METHODS: We extracted structured and unstructured data from the EHRs of participants of the Estonian Biobank (EstBB) and evaluated different formats of input data to understand how this continuously growing dataset should be prepared for best prediction. The utility of the EHR database for finding blood- and urine-based biomarkers for IS was demonstrated by applying different analytical and machine learning (ML) methods. RESULTS: Several early trends in common clinical laboratory parameter changes (set of red blood indices, lymphocyte/neutrophil ratio, etc.) were established for IS prediction. The developed ML models predicted the future occurrence of IS with very high accuracy and Random Forests was proved as the most applicable method to EHR data. CONCLUSIONS: We conclude that the EHR database and the risk factors uncovered are valuable resources in screening the population for risk of IS as well as constructing disease risk scores and refining prediction models for IS by ML.

A Survey of End-to-End Driving: Architectures and Training Methods.

Autonomous driving is of great interest to industry and academia alike. The use of machine learning approaches for autonomous driving has long been studied, but mostly in the context of perception. In this article, we take a deeper look on the so-called end-to-end approaches for autonomous driving, where the entire driving pipeline is replaced with a single neural network. We review the learning methods, input and output modalities, network architectures, and evaluation schemes in end-to-end driving literature. Interpretability and safety are discussed separately, as they remain challenging for this approach. Beyond providing a comprehensive overview of existing methods, we conclude the review with an architecture that combines the most promising elements of the end-to-end autonomous driving systems.

Perspective Taking in Deep Reinforcement Learning Agents.

Perspective taking is the ability to take into account what the other agent knows. This skill is not unique to humans as it is also displayed by other animals like chimpanzees. It is an essential ability for social interactions, including efficient cooperation, competition, and communication. Here we present our progress toward building artificial agents with such abilities. We implemented a perspective taking task inspired by experiments done with chimpanzees. We show that agents controlled by artificial neural networks can learn via reinforcement learning to pass simple tests that require some aspects of perspective taking capabilities. We studied whether this ability is more readily learned by agents with information encoded in allocentric or egocentric form for both their visual perception and motor actions. We believe that, in the long run, building artificial agents with perspective taking ability can help us develop artificial intelligence that is more human-like and easier to communicate with.

ViraMiner: Deep learning on raw DNA sequences for identifying viral genomes in human samples.

Despite its clinical importance, detection of highly divergent or yet unknown viruses is a major challenge. When human samples are sequenced, conventional alignments classify many assembled contigs as "unknown" since many of the sequences are not similar to known genomes. In this work, we developed ViraMiner, a deep learning-based method to identify viruses in various human biospecimens. ViraMiner contains two branches of Convolutional Neural Networks designed to detect both patterns and pattern-frequencies on raw metagenomics contigs. The training dataset included sequences obtained from 19 metagenomic experiments which were analyzed and labeled by BLAST. The model achieves significantly improved accuracy compared to other machine learning methods for viral genome classification. Using 300 bp contigs ViraMiner achieves 0.923 area under the ROC curve. To our knowledge, this is the first machine learning methodology that can detect the presence of viral sequences among raw metagenomic contigs from diverse human samples. We suggest that the proposed model captures different types of information of genome composition, and can be used as a recommendation system to further investigate sequences labeled as "unknown" by conventional alignment methods. Exploring these highly-divergent viruses, in turn, can enhance our knowledge of infectious causes of diseases.

Multiagent cooperation and competition with deep reinforcement learning.

Evolution of cooperation and competition can appear when multiple adaptive agents share a biological, social, or technological niche. In the present work we study how cooperation and competition emerge between autonomous agents that learn by reinforcement while using only their raw visual input as the state representation. In particular, we extend the Deep Q-Learning framework to multiagent environments to investigate the interaction between two learning agents in the well-known video game Pong. By manipulating the classical rewarding scheme of Pong we show how competitive and collaborative behaviors emerge. We also describe the progression from competitive to collaborative behavior when the incentive to cooperate is increased. Finally we show how learning by playing against another adaptive agent, instead of against a hard-wired algorithm, results in more robust strategies. The present work shows that Deep Q-Networks can become a useful tool for studying decentralized learning of multiagent systems coping with high-dimensional environments.