Search

Nowadays, deep neural networks based software have been widely applied in many areas including safety-critical areas such as traffic control, medical diagnosis and malware detection, etc. However, the software engineering techniques, which are supposed to guarantee the functionality, safety as well as fairness, are not well studied. For example, some serious crashes of DNN based autonomous cars have been reported. These crashes could have been avoided if these DNN based software were well tested. Traditional software testing, debugging or repairing techniques do not work well on DNN based software because there is no control flow, data flow or AST(Abstract Syntax Tree) in deep neural networks. Proposing software engineering techniques targeted on DNN based software are imperative. In this thesis, we first introduced the development of SE(Software Engineering) for AI(Artificial Intelligence) area and how our works have influenced the advancement of this new area. Then we summarized related works and some important concepts in SE for AI area. Finally, we discussed four important works of ours. Our first project DeepTest is one of the first few papers proposing systematic software testing techniques for DNN based software. We proposed neuron coverage guided image synthesis techniques for DNN based autonomous cars and leveraged domain specific metamorphic relation to generate oracle for new generated test cases to automatically test DNN based software. We applied DeepTest to testing three top performing self-driving car models in Udacity self-driving car challenge and our tool has identified thousands of erroneous behaviors that may lead to potential fatal crash. In DeepTest project, we found that the natural variation such as spatial transformations or rain/fog effects have led to problematic corner cases for DNN based self-driving cars. In the follow-up project DeepRobust, we studied per-point robustness of deep neural network under natural variation. We found that for a DNN model, some specific weak points are more likely to cause erroneous outputs than others under natural variation. We proposed a white-box approach and a black-box approach to identify these weak data points. We implemented and evaluated our approaches on 9 DNN based image classifiers and 3 DNN based self-driving car models. Our approaches can successfully detect weak points with good precision and recall for both DNN based image classifiers and self-driving cars. Most of existing works in SE for AI area including our DeepTest and DeepRobust focus on instance-wise errors, which are single inputs that result in a DNN model's erroneous outputs. Different from instance-wise errors, group-level errors reflect a DNN model's weak performance on differentiating among certain classes or inconsistent performance across classes. This type of errors is very concerning since it has been found to be related to many real-world notorious errors without malicious attackers. In our third project DeepInspect, we first introduced the group-level errors for DNN based software and categorized them into confusion errors and bias errors based on real-world reports. Then we proposed neuron coverage based distance metric to detect group-level errors for DNN based software without requiring labels. We applied DeepInspect to testing 8 pretrained DNN models trained in 6 popular image classification datasets, including three adversarial trained models. We showed that DeepInspect can successfully detect group-level violations for both single-label and multi-label classification models with high precision. As a follow-up and more challenging research project, we proposed five WR(weighted regularization) techniques to repair group-level errors for DNN based software. These five different weighted regularization techniques function at different stages of retraining or inference of DNNs including input phase, layer phase, loss phase and output phase. We compared and evaluated these five different WR techniques in both single-label and multi-label classifications including five combinations of four DNN architectures on four datasets. We showed that WR can effectively fix confusion and bias errors and these methods all have their pros, cons and applicable scenario. All our four projects discussed in this thesis have solved important problems in ensuring the functionality, safety as well as fairness for DNN based software and had significant influence in the advancement of SE for AI area.

With the proliferation of multi-site neuroimaging studies, there is a greater need for handling non-biological variance introduced by differences in MRI scanners and acquisition protocols. Such unwanted sources of variation, which we refer to as “scanner effects”, can hinder the detection of imaging features associated with clinical covariates of interest and cause spurious findings. In this paper, we investigate scanner effects in two large multi-site studies on cortical thickness measurements, across a total of 11 scanners. We propose a set of general tools for visualizing and identifying scanner effects that are generalizable to other modalities. We then propose to use ComBat, a technique adopted from the genomics literature and recently applied to diffusion tensor imaging data, to combine and harmonize cortical thickness values across scanners. We show that ComBat removes unwanted sources of scan variability while simultaneously increasing the power and reproducibility of subsequent statistical analyses. We also show that ComBat is useful for combining imaging data with the goal of studying life-span trajectories in the brain.

Alzheimer's Disease (AD) is the most common neurodegenerative disease in elderly people, and current drugs, unfortunately, do not represent yet a cure but only slow down its progression. This is explained, at least in part, because the understanding of the neurodegenerative process is still incomplete, being sometimes mistaken, particularly at the first steps of the illness, with the natural aging process. A better identification of how the functional activity deteriorates is thus crucial to develop new and more effective treatments. Sparse inverse covariance estimates (SICE) have been recently employed for deriving functional connectivity patterns from Positron Emission Tomography (PET) of brains affected by Alzheimer's Disease. SICE, unlike the traditional covariance methods, allows to analyze the interdependencies between brain regions factoring out the influence of others. To analyze the effects of the illness, connectivity patterns of brains affected by AD are compared with those obtained for control groups. These comparisons are, however, carried out for binary (undirected and unweighted) adjacency matrices with the same number of arcs. Additionally, the effect of the number of subjects employed or the validity of the regularization parameter used to compute the SICE have been not hitherto analyzed. In this paper, we delve into the construction of connectivity patterns from PET using SICE. In particular, we describe the effect that the number of subjects employed has on the results and identify, based on the reconstruction error of linear regression systems, a range of valid values for the regularization parameter. The amount of arcs is also proved as a discriminant value, and we show that it is possible to pass from unweighted (binary) to weighted adjacency matrices, where the weight of a connection corresponding to the existence of a relationship between two brain areas can be correlated to the persistence of this relationship when computed for different values of the regularization parameter and sets of subjects. Finally, network measures are computed for the connectivity patterns confirming that SICE may be particularly apt for assessing the efficiency of drugs, since it produces reliable brain connectivity models with small sample sizes, and that connectivity patterns affected by AD seem much less segregated, reducing the small-worldness.

High-throughput gene expression data are often obtained from pure or complex (heterogeneous) biological samples. In the latter case, data obtained are a mixture of different cell types and the heterogeneity imposes some difficulties in the analysis of such data. In order to make conclusions on gene expresssion data obtained from heterogeneous samples, methods such as microdissection and flow cytometry have been employed to physically separate the constituting cell types. However, these manual approaches are time consuming when measuring the responses of multiple cell types simultaneously. In addition, exposed samples, on many occasions, end up being contaminated with external perturbations and this may result in an altered yield of molecular content. In this paper, we model the heterogeneous gene expression data using a Bayesian framework, treating the cell type proportions and the cell-type specific expressions as the parameters of the model. Specifically, we present a novel sequential Monte Carlo (SMC) sampler for estimating the model parameters by approximating their posterior distributions with a set of weighted samples. The SMC framework is a robust and efficient approach where we construct a sequence of artificial target (posterior) distributions on spaces of increasing dimensions which admit the distributions of interest as marginals. The proposed algorithm is evaluated on simulated datasets and publicly available real datasets, including Affymetrix oligonucleotide arrays and national center for biotechnology information (NCBI) gene expression omnibus (GEO), with varying number of cell types. The results obtained on all datasets show a superior performance with an improved accuracy in the estimation of cell type proportions and the cell-type specific expressions, and in addition, more accurate identification of differentially expressed genes when compared to other widely known methods for blind decomposition of heterogeneous gene expression data such as Dsection and the nonnegative matrix factorization (NMF) algorithms. MATLAB implementation of the proposed SMC algorithm is available to download at https://github.com/moyanre/smcgenedeconv.git.