Dynamic stopping in a calibration-less P300 speller

Recent progress in machine learning has been mainly due to the availability of large amounts of annotated data used for training complex models with deep architectures. Annotating this training data becomes burdensome and creates a major bottleneck in maintaining machine-learning databases. Moreover, these trained models fail to generalize to new categories or new varieties of the same categories. This is because new categories or new varieties have data distribution different from the training data distribution. To tackle these problems, this thesis proposes to develop a family of transfer-learning techniques that can deal with different training (source) and testing (target) distributions with the assumption that the availability of annotated data is limited in the testing domain. This is done by using the auxiliary data-abundant source domain from which useful knowledge is transferred that can be applied to data-scarce target domain. This transferable knowledge serves as a prior that biases target-domain predictions and prevents the target-domain model from overfitting. Specifically, we explore structural priors that encode relational knowledge between different data entities, which provides more informative bias than traditional priors. The choice of the structural prior depends on the information availability and the similarity between the two domains. Depending on the domain similarity and the information availability, we divide the transfer learning problem into four major categories and propose different structural priors to solve each of these sub-problems. This thesis first focuses on the unsupervised-domain-adaptation problem, where we propose to minimize domain discrepancy by transforming labeled source-domain data to be close to unlabeled target-domain data. For this problem, the categories remain the same across the two domains and hence we assume that the structural relationship between the source-domain samples is carried over to the target domain. Thus, graph or hyper-graph is constructed as the structural prior from both domains and a graph/hyper-graph matching formulation is used to transform samples in the source domain to be closer to samples in the target domain. An efficient optimization scheme is then proposed to tackle the time and memory inefficiencies associated with the matching problem. The few-shot learning problem is studied next, where we propose to transfer knowledge from source-domain categories containing abundantly labeled data to novel categories in the target domain that contains only few labeled data. The knowledge transfer biases the novel category predictions and prevents the model from overfitting. The knowledge is encoded using a neural-network-based prior that transforms a data sample to its corresponding class prototype. This neural network is trained from the source-domain data and applied to the target-domain data, where it transforms the few-shot samples to the novel-class prototypes for better recognition performance. The few-shot learning problem is then extended to the situation, where we do not have access to the source-domain data but only have access to the source-domain class prototypes. In this limited information setting, parametric neural-network-based priors would overfit to the source-class prototypes and hence we seek a non-parametric-based prior using manifolds. A piecewise linear manifold is used as a structural prior to fit the source-domain-class prototypes. This structure is extended to the target domain, where the novel-class prototypes are found by projecting the few-shot samples onto the manifold. Finally, the zero-shot learning problem is addressed, which is an extreme case of the few-shot learning problem where we do not have any labeled data in the target domain. However, we have high-level information for both the source and target domain categories in the form of semantic descriptors. We learn the relation between the sample space and the semantic space, using a regularized neural network so that classification of the novel categories can be carried out in a common representation space. This same neural network is then used in the target domain to relate the two spaces. In case we want to generate data for the novel categories in the target domain, we can use a constrained generative adversarial network instead of a traditional neural network. Thus, we use structural priors like graphs, neural networks and manifolds to relate various data entities like samples, prototypes and semantics for these different transfer learning sub-problems. We explore additional post-processing steps like pseudo-labeling, domain adaptation and calibration and enforce algorithmic and architectural constraints to further improve recognition performance. Experimental results on standard transfer learning image recognition datasets produced competitive results with respect to previous work. Further experimentation and analyses of these methods provided better understanding of machine learning as well.

We present an automated method to track and identify neurons in C. elegans, called "fast Deep Learning Correspondence" or fDLC, based on the transformer network architecture. The model is trained once on empirically derived synthetic data and then predicts neural correspondence across held-out real animals via transfer learning. The same pre-trained model both tracks neurons across time and identifies corresponding neurons across individuals. Performance is evaluated against hand-annotated datasets, including NeuroPAL [1]. Using only position information, the method achieves 80.0% accuracy at tracking neurons within an individual and 65.8% accuracy at identifying neurons across individuals. Accuracy is even higher on a published dataset [2]. Accuracy reaches 76.5% when using color information from NeuroPAL. Unlike previous methods, fDLC does not require straightening or transforming the animal into a canonical coordinate system. The method is fast and predicts correspondence in 10 ms making it suitable for future real-time applications.

Caveolae are plasma membrane invaginations whose formation requires caveolin-1 (Cav1), the adaptor protein polymerase I, and the transcript release factor (PTRF or CAVIN1). Caveolae have an important role in cell functioning, signaling, and disease. In the absence of CAVIN1/PTRF, Cav1 forms non-caveolar membrane domains called scaffolds. In this work, we train machine learning models to automatically distinguish between caveolae and scaffolds from single molecule localization microscopy (SMLM) data. We apply machine and deep learning algorithms to discriminate biological structures from SMLM data. Our work is the first that is leveraging machine and deep learning approaches to automatically identifying biological structures from SMLM data. In particular, we develop and compare three binary classification methods to identify whether or not a given 3D cluster of Cav1 proteins is a caveolae. The first uses a random forest classifier applied to 28 hand-picked features, the second uses a convolutional neural net (CNN) applied to a projection of the point clouds onto three planes, and the third uses a PointNet model, a recent development that can directly take point clouds as its input. We validate our methods on a dataset of super-resolution microscopy images of PC3 prostate cancer cells labeled for Cav1. Specifically, we have images from two cell populations: 10 PC3 and 10 CAVIN1/PTRF-transfected PC3 cells (PC3-PTRF cells) that form caveolae. We obtained a balanced set of 1714 different cellular structures. Our results show that both the random forest on hand selected features and the deep learning approach achieve high accuracy in distinguishing the intrinsic features of the caveolae and non-caveolae biological structures. More specifically, both random forest and deep CNN classifiers achieve classification accuracy reaching 94% on our test set, while the PointNet model only reached 83% accuracy. We also discuss the pros and cons of the different approaches. Additional information about individual files is in the accompanying CSV file (item_metadata.csv). This dataset was originally deposited in the Simon Fraser University institutional repository.

The determination of transcriptional regulatory networks is key to the understanding of biological systems. However, the experimental determination of transcriptional regulatory networks in the laboratory remains difficult and time-consuming, while current computational methods to infer these networks (typically from gene-expression data) achieve only modest accuracy. The latter can be attributed in part to the limitations of a single-organism approach. Computational biology has long used comparative and, more generally, evolutionary approaches to extend the reach and accuracy of its analyses. We therefore use an evolutionary approach to the inference of regulatory networks, which enables us to study evolutionary models for these networks as well as to improve the accuracy of inferred networks. Since the regulatory networks evolve along with the genomes, we consider that the regulatory networks for a family of organisms are related to each other through the same phylogenetic tree. These relationships contain information that can be used to improve the accuracy of inferred networks. Advances in the study of evolution of regulatory networks provide evidence to establish evolutionary models for regulatory networks, which is an important component of our evolutionary approach. We use two network evolutionary models, a basic model that considers only the gains and losses of regulatory connections during evolution, and an extended model that also takes into account the duplications and losses of genes. With the network evolutionary models, we design refinement algorithms to make use of the phylogenetic relationships to refine noisy regulatory networks for a family of organisms. These refinement algorithms include: RefineFast and RefineML, which are two-step iterative algorithms, and ProPhyC and ProPhyCC, which are based on a probabilistic phylogenetic model. For each algorithm we first design it with the basic network evolutionary model and then generalize it to the extended evolutionary model. All these algorithms are computationally efficient and are supported by extensive experimental results showing that they yield substantial improvement in the quality of the input noisy networks. In particular, ProPhyC and ProPhyCC further improve the performance of RefineFast and RefineML. Besides the four refinement algorithms mentioned above, we also design an algorithm based on transfer learning theory called tree transfer learning (TTL). TTL differs from the previous four refinement algorithms in the sense that it takes the gene-expression data for the family of organisms as input, instead of their inferred noisy networks. TTL then learns the network structures for all the organisms at once, meanwhile taking advantage of the phylogenetic relationships. Although this approach outperforms an inference algorithm used alone, it does not perform better than ProPhyC, which indicates that the ProPhyC framework makes good use of the phylogenetic information.

This paper discusses a novel hybrid approach to text classification that integrates a machine learning algorithm along with DistilBERT, a pre-trained deep learning framework for natural language processing, offers a base model fine-tuned on Indian Banking News-Events with a rule-based method that is used by filtering false positives and dealing with false negatives to enhance the results given by the previous classifier. The major benefit is that by incorporating unique rules for certain chaotic or overlapping categories that have not been effectively trained, the system can be quickly fine-tuned. This research also compares the effectiveness of state-of-art deep contextual language representation DistilBERT used in our proposed hybrid model with the most preferred context independent language representation TFIDF, on supervised learning of the classification of multiclass Banking News-Events. Both representations are fed into the machine learning classifiers, Logistic Regression, Linear SVC, Decision Tree, and Random Forest. The findings indicate that DistilBERT can better transfer generic domain knowledge to other domains as compared to the baseline TFIDF and results into higher accuracy with Random Forest. The result is further fed into the Rules based method as per our proposed model. And, our proposed hybrid model resulted into improved accuracy.

This dissertation addresses two significant issues in the effective application of machine learning algorithms and models for the physical and engineering sciences. The first is the broad challenge of automated modeling of data across different processes in a physical system. The second is the dilemma of obtaining insightful interpretations on the relationships between the inputs and outcome of a system as inferred from complex, black box machine learning models.Automated Geometric Shape Deviation Modeling for Additive Manufacturing SystemsAdditive manufacturing systems possess an intrinsic capability for one-of-a-kind manufacturing of a vast variety of shapes across a wide spectrum of processes. One major issue in AM systems is geometric accuracy control for the inevitable shape deviations that arise in AM processes. Current effective approaches for shape deviation control in AM involve the specification of statistical or machine learning deviation models for additively manufactured products. However, this task is challenging due to the constraints on the number of test shapes that can be manufactured in practice, and limitations on user efforts that can be devoted for learning deviation models across different shape classes and processes in an AM system. We develop an automated, Bayesian neural network methodology for comprehensive shape deviation modeling in an AM system. A fundamental innovation in this machine learning method is our new and connectable neural network structures that facilitate the transfer of prior knowledge and models on deviations across different shape classes and AM processes. Several case studies on in-plane and out-of-plane deviations, regular and free-form shapes, and different settings of lurking variables serve to validate the power and broad scope of our methodology, and its potential to advance high-quality manufacturing in an AM system.Interpretable Machine LearningMachine learning algorithms and models constitute the dominant set of predictive methods for a wide range of complex, real-world processes. However, interpreting what such methods effectively infer from data is difficult in general. This is because their typical black box natures possess a limited ability to directly yield insights on the underlying relationships between inputs and the outcome for a process. We develop methodologies based on new predictive comparison estimands that effectively enable one to ``mine’’ machine learning models, in the sense of (a) interpreting their inferred associations between inputs and/or functional forms of inputs with the outcome, (b) identifying the inputs that they effectively consider relevant, and (c) interpreting the inferred conditional and two-way associations of the inputs with the outcome. We establish Fisher consistent estimators, and their corresponding standard errors, for our new estimands under a condition on the inputs' distributions. The significance of our predictive comparison methodology is demonstrated with a wide range of simulation and case studies that involve Bayesian additive regression trees, neural networks, and support vector machines. Our extended study of interpretable machine learning for AM systems demonstrates how our method can contribute to smarter advanced manufacturing systems, especially as current machine learning methods for AM are lacking in their ability to yield meaningful engineering knowledge on AM processes.

We introduce the novel problem of inter-robot transfer learning for perceptual classification of objects, where multiple heterogeneous robots communicate and transfer learned object models consisting of a fusion of multiple object properties. Unlike traditional transfer learning, there can be severe differences in the data distributions, resulting from differences in sensing, sensory processing, or even representations, that each robot uses to learn. Furthermore, only some properties may overlap between the two robots. We show that in such cases, the abstraction of raw sensory data into an intermediate representation can be used not only to aid learning, but also the transfer of knowledge. Further, we utilize statistical metrics, learned during an interactive process where the robots jointly explore the environment, to determine which underlying properties are shared between the robots. We demonstrate results in a visual classification task where objects are represented via a combination of properties derived from different modalities: color, texture, shape, and size. Using our methods, two heterogeneous robots utilizing different sensors and representations are able to successfully transfer support vector machine (SVM) classifiers among each other, resulting in speedups during learning.

Active Learning is a machine learning and data mining technique that selects the most informative samples for labeling and uses them as training data; it is especially useful when there are large amount of unlabeled data and labeling them is expensive. Recently, batch-mode active learning, where a set of samples are selected concurrently for labeling, based on their collective merit, has attracted a lot of attention. The objective of batch-mode active learning is to select a set of informative samples so that a classifier learned on these samples has good generalization performance on the unlabeled data. Most of the existing batch-mode active learning methodologies try to achieve this by selecting samples based on certain criteria. In this article we propose a novel criterion which achieves good generalization performance of a classifier by specifically selecting a set of query samples that minimize the difference in distribution between the labeled and the unlabeled data, after annotation. We explicitly measure this difference based on all candidate subsets of the unlabeled data and select the best subset. The proposed objective is an NP-hard integer programming optimization problem. We provide two optimization techniques to solve this problem. In the first one, the problem is transformed into a convex quadratic programming problem and in the second method the problem is transformed into a linear programming problem. Our empirical studies using publicly available UCI datasets and two biomedical image databases demonstrate the effectiveness of the proposed approach in comparison with the state-of-the-art batch-mode active learning methods. We also present two extensions of the proposed approach, which incorporate uncertainty of the predicted labels of the unlabeled data and transfer learning in the proposed formulation. In addition, we present a joint optimization framework for performing both transfer and active learning simultaneously unlike the existing approaches of learning in two separate stages, that is, typically, transfer learning followed by active learning. We specifically minimize a common objective of reducing distribution difference between the domain adapted source, the queried and labeled samples and the rest of the unlabeled target domain data. Our empirical studies on two biomedical image databases and on a publicly available 20 Newsgroups dataset show that incorporation of uncertainty information and transfer learning further improves the performance of the proposed active learning based classifier. Our empirical studies also show that the proposed transfer-active method based on the joint optimization framework performs significantly better than a framework which implements transfer and active learning in two separate stages.

Improved P300 speller performance using electrocorticography, spectral features, and natural language processing

Although reinforcement learning (RL) has been successfully deployed in a variety of tasks, learning speed remains a fundamental problem for applying RL in complex environments. Transfer learning aims to ameliorate this shortcoming by speeding up learning through the adaptation of previously learned behaviors in similar tasks. Transfer techniques often use an inter-task mapping, which determines how a pair of tasks are related. Instead of relying on a hand-coded inter-task mapping, this paper proposes a novel transfer learning method capable of autonomously creating an inter-task mapping by using a novel combination of sparse coding, sparse projection learning and sparse Gaussian processes. We also propose two new transfer algorithms (TrLSPI and TrFQI) based on least squares policy iteration and fitted-Q-iteration. Experiments not only show successful transfer of information between similar tasks, inverted pendulum to cart pole, but also between two very different domains: mountain car to cart pole. This paper empirically shows that the learned inter-task mapping can be successfully used to (1) improve the performance of a learned policy on a fixed number of environmental samples, (2) reduce the learning times needed by the algorithms to converge to a policy on a fixed number of samples, and (3) converge faster to a near-optimal policy given a large number of samples.

In recent years, discriminative learning methods have widely been used in various areas of Natural Language Processing (NLP). These methods achieve the best performance, when the set of training and testing samples have the same distribution. However, in many applications of NLP, the lack of labeled datasets for some domains is a serious challenge. In such conditions, we need to develop a model based on domains with rich labeled instances and apply it to the domain with no labeled instances. In this research, a method for sentiment classification of opinions into positive and negative groups, which represent the users' feelings, is offered based on multi-source transfer learning. The proposed method here employs Spectral Feature Alignment algorithm to adapt different domains. Furthermore, according to the Majority Voting, accuracy is assigned to classifications trained on different domains based on the Majority Voting Error. Ultimately, decisions are made for each classification based on the calculated error. The Amazon datasets for four different categories, each of which contains 1000 positive and 1000 negative samples, are exploited to train the proposed model. Meanwhile, each category includes unlabeled samples that are used to select pivot features. The accuracy values of 85.5%, 86.4%, 83.5% and 90.1% obtained for Electronics, DVD, Books and Kitchen domains respectively, show the effectiveness of the proposed method compared with similar methods.

Reinvigorated by algorithmic developments, faster hardware and large data sets, machine learning is pervading many aspects of chemistry. We present two examples from our recent studies, one in the area of drug discovery, the other focused on protein spectroscopy. In the first, we consider pharmaceutical lead discovery as active search in a space of labelled graphs [1]. We extend a recent data-driven adaptive Markov chain approach, and evaluate it on a focused drug design problem, where we search for an antagonist of an 훂v integrin, the target protein that belongs to a group of Arg-Gly-Asp integrin receptors. In the second example, we present a novel machine learning protocol that uses a few key structural descriptors to predict amide I IR spectra of proteins and agrees well with experiment [2]. Its transferability enabled us to distinguish protein secondary structures, probe atomic structure variations with temperature, and monitor protein folding.[1] Oglic, D., Oatley, S.A., Macdonald, S.J.F., McInally, T., Garnett, R., Hirst, J.D. & Gartner, T. Active search for computer-aided drug design. Mol. Inf., 2018, 37, 1700130. https://doi.org/10.1002/minf.201700130[2] Ye, S., Zhong, K., Zhang, J., Hu, W., Hirst, J.D., Zhang, G., Mukamel, S., Jiang, J. A transferable machine learning protocol for predicting protein amide-I infrared spectra. J. Am. Chem. Soc., 2020, 142, 19071. https://doi.org/10.1021/jacs.0c06530

Robots had a great impact on the manufacturing industry ever since the early seventies when companies such as KUKA and ABB started deploying their first industrial robots. These robots merely performed very specific tasks in specific ways within well-defined environments. Still, they proved to be very useful as they could exceed human performance at these tasks. However, in order to enable robots to enter our daily life, they need to become more versatile and need to operate in much less structured environments. This thesis is partly devoted to stretching these limitations by means of learning, namely imitation learning (IL) and inverse reinforcement learning (IRL). Reinforcement learning (RL) is a powerful approach to enable robots to solve a task in an unknown environment. The practitioner describes a desired behavior by specifying a reward function and the robot autonomously interacts with the environment in order to find a control policy that generates high accumulated reward. However, RL is not suitable for teaching new tasks by non-experts because specifying appropriate reward functions can be difficult. Demonstrating the desired behavior is often easier for non-experts. Imitation learning can be used in order to enable the robot to reproduce the demonstrations. However, without explicitly inferring and modeling the intentions of the demonstrations, it can become difficult to solve the task for unseen situations. Inverse reinforcement learning (IRL) therefore aims to infer a reward function from the demonstrations, such that optimizing this reward function yields the desired behavior even for different situations. This thesis introduces a unifying approach to solve the inverse reinforcement learning problem in the same way as the reinforcement learning problem. This is achieved by framing both problems as information projection problems, i.e., we strive to minimize the relative entropy between a probabilistic model of the robot behavior and a given desired distribution. Furthermore, a trust region on the robot behavior is used to stabilize the optimization. For inverse reinforcement learning, the desired distribution is implicitly given by the expert demonstrations. The resulting optimization can be efficiently solved using state-of-the-art reinforcement learning methods. For reinforcement learning, the log-likelihood of the desired distribution is given by the reward function. The resulting optimization problem corresponds to a standard reinforcement learning formulation, except for an additional objective of maximizing the entropy of the robot behavior. This entropy objective adds little overhead to the optimization, but can lead to better exploration and more diversified policies. Trust-region I-projections are not only useful for training robots, but can also be applied to other machine learning problems. I-projections are typically used for variational inference, in order to approximate an intractable distribution by a simpler model. However, the resulting optimization problems are usually optimized based on stochastic gradient descent which often suffers from high variance in the gradient estimates. As trust-region I-projections where shown to be effective for reinforcement learning and inverse reinforcement learning, this thesis also explores their use for variational inference. More specifically, trust-region I-projections are investigated for the problem of approximating an intractable distribution by a Gaussian mixture model (GMM) with an adaptive number of components. GMMs are highly desirable for variational inference because they can yield arbitrary accurate approximations while inference from GMMs is still relatively cheap. In order to make learning the GMM feasible, we derive a lower bound that enables us to decompose the objective function. The optimization can then be performed by iteratively updating individual components using a technique from reinforcement learning. The resulting method is capable of learning approximations of significantly higher quality than existing variational inference methods. Due to the similarity of the underlying optimization problems, the insights gained from our variational inference method are also useful for IL and IRL. Namely, a similar lower bound can be applied also for the I-projection formulation of imitation learning. However, whereas for variational inference the lower bound serves to decompose the objective function, for imitation learning it allows us to provide a reward signal to the robot that does not depend on its behavior. Compared to reward functions that are relative to the current behavior of the robot---which are typical for popular adversarial methods---behavior-independent reward functions have the advantages that we can show convergence even for greedy optimization. Furthermore, behavior-independent reward functions solve the inverse reinforcement learning problem, thereby closing the gap between imitation learning and IRL. However, algorithms derived from our non-adversarial formulation are actually very similar to existing AIL methods, and we can even show that adversarial inverse reinforcement learning (AIRL) is indeed an instance of our formulation. AIRL was derived from an adversarial formulation, and we point out several problems of that derivation. In contrast, we show that AIRL can be straightforwardly derived from out non-adversarial formulation. Furthermore, we demonstrate that the non-adversarial formulation can be also used to derive novel algorithms by presenting a non-adversarial method for offline imitation learning.

Recently, the ransomware attack poses a serious threat that targets a wide range of organizations and individuals for financial gain. So, there is a real need to initiate more innovative methods that are capable of proactively detect and prevent this type of attack. Multiple approaches were innovated to detect attacks using different techniques. One of these techniques is machine learning techniques which provide reasonable results, in most attack detection systems. In the current article, different machine learning techniques are tested to analyze its ability in a detection ransomware attack. The top 1000 features extracted from raw byte with the use of gain ratio as a feature selection method. Three different classifiers (Decision Tree (J48), Random Forest, radial basis function RBF network) available in WEKA based machine learning tool are evaluated to achieve significant detection accuracy of ransomware. The result shows that Random Forest gave the best detection accuracy almost around 98%.

This work explores better adaptation methods to low-resource languages using an external language model (LM) under the framework of transfer learning. We first build a language-independent ASR system in a unified sequence-to-sequence (S2S) architecture with a shared vocabulary among all languages. During adaptation, we perform LM fusion transfer, where an external LM is integrated into the decoder network of the attention-based S2S model in the whole adaptation stage, to effectively incorporate linguistic context of the target language. We also investigate various seed models for transfer learning. Experimental evaluations using the IARPA BABEL data set show that LM fusion transfer improves performances on all target five languages compared with simple transfer learning when the external text data is available. Our final system drastically reduces the performance gap from the hybrid systems.