Browsing by Subject "Artificial intelligence"
Now showing 1 - 20 of 54
- Results Per Page
- Sort Options
Item A modular attention hypothesis for modeling visuomotor behaviors(2021-07-24) Zhang, Ruohan; Ballard, Dana H. (Dana Harry), 1946-; Hayhoe, Mary; Stone, Peter; Huth, Alexander; Dayan, PeterIn this dissertation, we explore the hypothesis that complex intelligent behaviors, in vivo, can be decomposed into modules, which are organized in hierarchies and executed in parallel. This organization is similar to a multiprocessing architecture in silico. Biological attention can be viewed as a "process manager" that manages information processing and multiple computations. In this work, we seek to understand and model this modular attention mechanism for humans in a range of behavioral settings. We explain this approach to understanding modular attention at three levels based on David Marr’s paradigm: the computation theory level, the representation and algorithm level, and the hardware implementation level. At the computation theory level, we propose that simple visuomotor behaviors can be broken down into modules that require attention for their execution. At the representation and algorithm level, we model human eye movements and actions in a variety of visuomotor tasks. We collect and publish a large-scale, high-quality dataset of eye movements and actions of humans playing Atari video games. We study the active vision problem by jointly modeling human eye movements and actions, and compare how humans and artificial learning agents play these video games differently. We then propose a modular reinforcement learning model for modeling human subjects’ navigation behaviors in a virtual-reality environment with multiple goals. We further develop a modular inverse reinforcement learning algorithm to efficiently estimate the subjective reward and discount factors associated with each behavioral goal. At the implementation level, we propose a theoretical neuronal communication model named gamma spike multiplexing that allows the cortex to perform multiple computations simultaneously without crosstalk. The model explains how the modular attention hypothesis might be implemented in the biological brain. The end goals of this work are to (1) build models to explain and predict observed human visuomotor behaviors and attention; (2) use these biologically inspired models to develop algorithms for better artificial learning systems.Item Active visual category learning(2011-05) Vijayanarasimhan, Sudheendra; Grauman, Kristen Lorraine, 1979-; Dhillon, Inderjit S.; Aggarwal, J K.; Mooney, Raymond J.; Torralba, AntonioVisual recognition research develops algorithms and representations to autonomously recognize visual entities such as objects, actions, and attributes. The traditional protocol involves manually collecting training image examples, annotating them in specific ways, and then learning models to explain the annotated examples. However, this is a rather limited way to transfer human knowledge to visual recognition systems, particularly considering the immense number of visual concepts that are to be learned. I propose new forms of active learning that facilitate large-scale transfer of human knowledge to visual recognition systems in a cost-effective way. The approach is cost-effective in the sense that the division of labor between the machine learner and the human annotators respects any cues regarding which annotations would be easy (or hard) for either party to provide. The approach is large-scale in that it can deal with a large number of annotation types, multiple human annotators, and huge pools of unlabeled data. In particular, I consider three important aspects of the problem: (1) cost-sensitive multi-level active learning, where the expected informativeness of any candidate image annotation is weighed against the predicted cost of obtaining it in order to choose the best annotation at every iteration. (2) budgeted batch active learning, a novel active learning setting that perfectly suits automatic learning from crowd-sourcing services where there are multiple annotators and each annotation task may vary in difficulty. (3) sub-linear time active learning, where one needs to retrieve those points that are most informative to a classifier in time that is sub-linear in the number of unlabeled examples, i.e., without having to exhaustively scan the entire collection. Using the proposed solutions for each aspect, I then demonstrate a complete end-to-end active learning system for scalable, autonomous, online learning of object detectors. The approach provides state-of-the-art recognition and detection results, while using minimal total manual effort. Overall, my work enables recognition systems that continuously improve their knowledge of the world by learning to ask the right questions of human supervisors.Item Adapting to unseen driving conditions using context-aware neural networksAbdulquddos, Suhaib; Miikkulainen, Risto; Tutum, Cem COne of the primary inhibitors to successful deployment of autonomous agents in real-world tasks such as driving is their poor ability to adapt to unseen conditions. Whereas a human might be able to deduce the best course of action when confronted with an unfamiliar set of conditions based on past experiences, artificial agents have difficulty performing in conditions that are significantly different from those in which they were trained. This thesis explores an approach in which the addition of a context module to a neural network is used to overcome the challenge of adapting to unseen conditions during evaluation. The approach is tested in the CARLA simulator wherein the torque and steering curves of a vehicle are modified during training and evaluation. Furthermore the agent is trained only on a track with a relatively large radius of curvature but is evaluated on a track with much sharper turns and the agent must learn to adapt its speed and steering during evaluation. Three different neural network architectures are used for these experiments, and their respective performances are compared: Context+Skill, Context only, Skill only. It is observed that when both performance and safety of agents behavior are considered, the context+skill network consistently outperforms both the skill only and the context only architectures. The results presented in this thesis indicate that the context aware approach is a promising step towards solving the generalization problem in the autonomous vehicle domain. Furthermore, this research presents a framework for comparing the generalization capabilities of various network architectures and approaches. It is posited that the context+skill neural network has the potential to advance the field of machine learning with regards to generalization in domains beyond just autonomous driving; that is, any domain where awareness of changing environment parameters can have a positive impact on performance.Item An artificial intelligence based automation solution for optimal scheduling of ELISA assays(2005-12-24) Ahuja, Manish; Koen, B. V.The thesis focuses on designing a hardware solution and building an optimal scheduling solution for automating ELISA assays used in clinical diagnostics. The hardware solution consists of a cylindrical robot running on a rail which integrates stand-alone instruments. The scheduling algorithm provides software solution which provides a framework of automation. Assays running on such an automated platform have unique requirements which differentiate their scheduling from ordinary job shop scheduling problems. The first difference is that an assay cannot wait at any stage once it has entered the system. Thus waiting time has to be zero. This is a rigid requirement, not found in job shop scheduling. The second difference is that a resource/machine can be used again and again after first processing has finished. This means an assay can return to the same machine after undergoing some operations at other machines. These two requirements and different needs of different assays make this problem fairly complicated. Two scheduling algorithms are developed–Earliest Fit and Exhaustive Enumeration. The former is used in industry and is non-optimal, while the latter yields optimal solution. Exhaustive Enumeration is a procedure in which all possible schedules are formed. This is followed by discarding duplicate schedule by using the technique of pattern recognition. Finally, the schedule, which is optimal, is selectedItem Automatic interpretation of loosely encoded knowledge(2006) Fan, James Junmin; Porter, Bruce, 1956-Knowledge is critical for a variety of artificial intelligence problems. A key challenge in using knowledge-based systems is how to align one's encoding with the idiosyncrasies in the existing knowledge base. We call such misalignments "loose speak". We found that loose-speak occurs frequently in knowledge base interactions with such regularity that it can be interpreted automatically by a machine. We created a loose-speak interpreter based on a unified approach that is capable of interpreting the different forms of loose speak, and we evaluated it through empirical studies in different domains and on different tasks.Item Autonomous intersection management(2009-12) Dresner, Kurt Mauro; Stone, Peter, 1971-; Porter, Bruce W.; Waller, S T.; Kuipers, Benjamin J.; Veloso, Manuela M.Artificial intelligence research is ushering in an era of sophisticated, mass-market transportation technology. While computers can fly a passenger jet better than a human pilot, people still face the dangerous yet tedious task of driving. Intelligent Transportation Systems (ITS) is the field focused on integrating information technology with vehicles and transportation infrastructure. Recent advances in ITS point to a future in which vehicles handle the vast majority of the driving task. Once autonomous vehicles become popular, interactions amongst multiple vehicles will be possible. Current methods of vehicle coordination will be outdated. The bottleneck for efficiency will no longer be drivers, but the mechanism by which those drivers' actions are coordinated. Current methods for controlling traffic cannot exploit the superior capabilities of autonomous vehicles. This thesis describes a novel approach to managing autonomous vehicles at intersections that decreases the amount of time vehicles spend waiting. Drivers and intersections in this mechanism are treated as autonomous agents in a multiagent system. In this system, agents use a new approach built around a detailed communication protocol, which is also a contribution of the thesis. In simulation, I demonstrate that this mechanism can significantly outperform current intersection control technology-traffic signals and stop signs. This thesis makes several contributions beyond the mechanism and protocol. First, it contains a distributed, peer-to-peer version of the protocol for low-traffic intersections. Without any requirement of specialized infrastructure at the intersection, such a system would be inexpensive and easy to deploy at intersections which do not currently require a traffic signal. Second, it presents an analysis of the mechanism's safety, including ways to mitigate some failure modes. Third, it describes a custom simulator, written for this work, which will be made publicly available following the publication of the thesis. Fourth, it explains how the mechanism is "backward-compatible" so that human drivers can use it alongside autonomous vehicles. Fifth, it explores the implications of using the mechanism at multiple proximal intersections. The mechanism, along with all available modes of operation, is implemented and tested in simulation, and I present experimental results that strongly attest to the efficacy of this approach.Item Autonomous qualitative learning of distinctions and actions in a developing agent(2010-08) Mugan, Jonathan William; Kuipers, Benjamin; Stone, Peter, 1971-; Ballard, Dana; Cohen, Leslie; Mooney, RaymondHow can an agent bootstrap up from a pixel-level representation to autonomously learn high-level states and actions using only domain general knowledge? This thesis attacks a piece of this problem and assumes that an agent has a set of continuous variables describing the environment and a set of continuous motor primitives, and poses a solution for the problem of how an agent can learn a set of useful states and effective higher-level actions through autonomous experience with the environment. There exist methods for learning models of the environment, and there also exist methods for planning. However, for autonomous learning, these methods have been used almost exclusively in discrete environments. This thesis proposes attacking the problem of learning high-level states and actions in continuous environments by using a qualitative representation to bridge the gap between continuous and discrete variable representations. In this approach, the agent begins with a broad discretization and initially can only tell if the value of each variable is increasing, decreasing, or remaining steady. The agent then simultaneously learns a qualitative representation (discretization) and a set of predictive models of the environment. The agent then converts these models into plans to form actions. The agent then uses those learned actions to explore the environment. The method is evaluated using a simulated robot with realistic physics. The robot is sitting at a table that contains one or two blocks, as well as other distractor objects that are out of reach. The agent autonomously explores the environment without being given a task. After learning, the agent is given various tasks to determine if it learned the necessary states and actions to complete them. The results show that the agent was able to use this method to autonomously learn to perform the tasks.Item Bottlenecks in big data analytics and AI applications and opportunities for improvement(2022-12-21) Richins, Daniel J.; Janapa Reddi, Vijay; John, Lizy Kurian; Wu, Carole-Jean; Julien, Christine L; Marculescu, DianaFrom shopping to social interaction, the related domains of big data analytics and artificial intelligence (AI) applications affect many aspects of our daily activities. Their success arises in part from their highly parallelizable compute, which allows them to process massive data sets in data centers, serve large numbers of users simultaneously, and perform almost innumerable simple calculations very quickly. Despite the success and ubiquity of big data analytics and AI, I show that the foundational principle of high performance in these paradigms—abundant and easily exploited parallel computation—has been pushed to the point where the limitations of parallel computing have come to dictate application performance. Using the industry benchmark TPCx-BB, I demonstrate that most of the compute is spent in code regions unable to fully utilize the available cores. In accordance with Amdahl’s law, overall performance is dictated by these less-parallel regions of compute. And in a data center deployment of an end-to-end AI application, the abundant parallelism of DNN inference is overshadowed by the non-parallel portions of the application pipeline: pre- and post-processing and inter-server communication. In a study of accelerated AI, I show that at a modest 8x compute speedup, performance improvement is completely halted by the limited storage bandwidth of just a handful of servers. Even within DNN inference itself, the demand for higher performance is pushing current hardware to its limits, to the point where DNN accuracy must sometimes be sacrificed for latency. To address the limitations at the boundaries of parallel computing in these domains, I propose solutions targeted to each domain. In big data analytics, I demonstrate that restricting big data software to a small subset of the available cores on each server can substantially improve performance and I propose a combined hardware/software solution called core packing that would extend these benefits (up to 20% latency reduction) to a wide range of big data applications. In data center AI applications, I demonstrate how an edge data center, carefully tailored to the specific behavior of accelerated AI applications, can accommodate up to 32x accelerated AI at 15% lower total cost of ownership than a comparable data center that does not tailor itself to the needs of the application. And within DNN inference, I show that an additional source of parallelism—between adjacent layers in the DNN graph—can be exploited to offer latency reductions up to 39%.Item Cognitive Science in technology(2010-12) Cabrera, Victoria Marrujo; Lewis, Kyle, 1961-; Ambler, TonyCognitive Science is an interdisciplinary field that ties together the curricula of liberal arts and technical fields of study. However, it is de-emphasized in technical undergraduate studies such as Engineering. Cognitive Science is essentially the study of the human mind and how humans process information. It is the study of human responses, thinking, and perception. Human behavior and a person’s reaction are undetermined, but it can be better understood. Although human behavior and interaction is a routine part of life, engineers are taught to decipher code and not how to decipher a human’s behavior. Cognitive Science affects all aspects in the work environment. Organizational practices can be improved by understanding common biases and motivational theories in people. Having a cognitive awareness of typical human behavior will help to promote improved communication and positive reactions from people in the workplace. Human behavior is inevitable in any field but more crucial in technical fields especially when there is lack of communication or ambiguous guidelines and definitions. In technical fields, miscommunication or ambiguity can be a matter of life or death. In many situations, miscommunication can drive ambiguity. Although some people are happy with flexible guidelines, others seek to have well defined expectations. How do people react in situations surrounding miscommunication or ambiguity? In both situations, some people create opportunities and others become a hindrance. Processes and procedures can be put in place to alleviate ambiguous situations, but human performance and psychological factors still play a role as well. Human error can result from psychological factors, but the environment can be improved to limit those factors. As with any situation, mishaps are still prone to happen. Although human error is preventable in most cases, it’s never completely unavoidable. Human error continues to be a deep-rooted cause that can lead to negative outcomes. As stated by Alexander Pope, “to err is human…” (Moncur). This paper will explore underlying human behavior in daily activities. By understanding common biases and motivational theories driving human behavior, one can address negative behavior in a technical field in order to create opportunities.Item Creating and utilizing symbolic representations of spatial knowledge using mobile robots(2008-08) Beeson, Patrick Foil, 1977-; Kuipers, BenjaminA map is a description of an environment allowing an agent--a human, or in our case a mobile robot--to plan and perform effective actions. From a single location, an agent’s sensors can not observe the whole structure of a complex, large environment. For this reason, the agent must build a map from observations gathered over time and space. We distinguish between large-scale space, with spatial structure larger than the agent’s sensory horizon, and small-scale space, with structure within the sensory horizon. We propose a factored approach to mobile robot map-building that handles qualitatively different types of uncertainty by combining the strengths of topological and metrical approaches. Our framework is based on a computational model of the human cognitive map; thus it allows robust navigation and communication within several different spatial ontologies. Our approach factors the mapping problem into natural sub-goals: building a metrical representation for local small-scale spaces; finding a topological map that represents the qualitative structure of large-scale space; and (when necessary) constructing a metrical representation for large-scale space using the skeleton provided by the topological map. The core contributions of this thesis are a formal description of the Hybrid Spatial Semantic Hierarchy (HSSH), a framework for both small-scale and large-scale representations of space, and an implementation of the HSSH that allows a robot to ground the largescale concepts of place and path in a metrical model of the local surround. Given metrical models of the robot’s local surround, we argue that places at decision points in the world can be grounded by the use of a primitive called a gateway. Gateways separate different regions in space and have a natural description at intersections and in doorways. We provide an algorithmic definition of gateways, a theory of how they contribute to the description of paths and places, and practical uses of gateways in spatial mapping and learning.Item Curriculum learning in reinforcement learning(2021-05-06) Narvekar, Sanmit Santosh; Stone, Peter, 1971-; Niekum, Scott; Mooney, Raymond; Brunskill, EmmaIn recent years, reinforcement learning (RL) has been increasingly successful at solving complex tasks. Despite these successes, one of the fundamental challenges is that many RL methods require large amounts of experience, and thus can be slow to train in practice. Transfer learning is a recent area of research that has been shown to speed up learning on a complex task by transferring knowledge from one or more easier source tasks. Most existing transfer learning methods treat this transfer of knowledge as a one-step process, where knowledge from all the sources are directly transferred to the target. However, for complex tasks, it may be more beneficial (and even necessary) to gradually acquire skills over multiple tasks in sequence, where each subsequent task requires and builds upon knowledge gained in a previous task. This idea is pervasive throughout human learning, where people learn complex skills gradually by training via a curriculum. The goal of this thesis is to explore whether autonomous reinforcement learning agents can also benefit by training via a curriculum, and whether such curricula can be designed fully autonomously. In order to answer these questions, this thesis first formalizes the concept of a curriculum, and the methodology of curriculum learning in reinforcement learning. Curriculum learning consists of 3 main elements: 1) task generation, which creates a suitable set of source tasks; 2) sequencing, which focuses on how to order these tasks into a curriculum; and 3) transfer learning, which considers how to transfer knowledge between tasks in the curriculum. This thesis introduces several methods to both create suitable source tasks and automatically sequence them into a curriculum. We show that these methods produce curricula that are tailored to the individual sensing and action capabilities of different agents, and show how the curricula learned can be adapted for new, but related target tasks. Together, these methods form the components of an autonomous curriculum design agent, that can suggest a training curriculum customized to both the unique abilities of each agent and the task in question. We expect this research on the curriculum learning approach will increase the applicability and scalability of RL methods by providing a faster way of training reinforcement learning agents, compared to learning tabula rasa.Item Data efficient reinforcement learning with off-policy and simulated data(2019-11-12) Hanna, Josiah Paul; Stone, Peter, 1971-; Niekum, Scott; Krähenbühl, Philipp; Sutton, RichardLearning from interaction with the environment -- trying untested actions, observing successes and failures, and tying effects back to causes -- is one of the first capabilities we think of when considering autonomous agents. Reinforcement learning (RL) is the area of artificial intelligence research that has the goal of allowing autonomous agents to learn in this way. Despite much recent success, many modern reinforcement learning algorithms are still limited by the requirement of large amounts of experience before useful skills are learned. Two possible approaches to improving data efficiency are to allow algorithms to make better use of past experience collected with past behaviors (known as off-policy data) and to allow algorithms to make better use of simulated data sources. This dissertation investigates the use of such auxiliary data by answering the question, "How can a reinforcement learning agent leverage off-policy and simulated data to evaluate and improve upon the expected performance of a policy?" This dissertation first considers how to directly use off-policy data in reinforcement learning through importance sampling. When used in reinforcement learning, importance sampling is limited by high variance that leads to inaccurate estimates. This dissertation addresses this limitation in two ways. First, this dissertation introduces the behavior policy gradient algorithm that adapts the data collection policy towards a policy that generates data that leads to low variance importance sampling evaluation of a fixed policy. Second, this dissertation introduces the family of regression importance sampling estimators which improve the weighting of already collected off-policy data so as to lower the variance of importance sampling evaluation of a fixed policy. In addition to evaluation of a fixed policy, we apply the behavior policy gradient algorithm and regression importance sampling to batch policy gradient policy improvement. In the case of regression importance sampling, this application leads to the introduction of the sampling error corrected policy gradient estimator that improves the data efficiency of batch policy gradient algorithms. Towards the goal of learning from simulated experience, this dissertation introduces an algorithm -- the grounded action transformation algorithm -- that takes small amounts of real world data and modifies the simulator such that skills learned in simulation are more likely to carry over to the real world. Key to this approach is the idea of local simulator modification -- the simulator is automatically altered to better model the real world for actions the data collection policy would take in states the data collection policy would visit. Local modification necessitates an iterative approach: the simulator is modified, the policy improved, and then more data is collected for further modification. Finally, in addition to examining them each independently, this dissertation also considers the possibility of combining the use of simulated data with importance sampled off-policy data. We combine these sources of auxiliary data by control variate techniques that use simulated data to lower the variance of off-policy policy value estimation. Combining these sources of auxiliary data allows us to introduce two algorithms -- weighted doubly robust bootstrap and model-based bootstrap -- for the problem of lower-bounding the performance of an untested policy.Item Deep learning empowers the next generation of seismic interpretation(2020-05-08) Shi, Yunzhi; Fomel, Sergey B.; Sen, Mrinal K; Krähenbühl, Philipp; Zeng, Hongliu; Ghattas, Omar NWith the ever developing data acquisition techniques, seismic processing deals with massive amount of high quality 3-D data with greater pressure to interpret the data more efficiently. Currently, seismic interpretation such as fault analysis and salt detection is a tedious, manual, and time-consuming process. Modern interpretive tools still rely on interpreter while only utilizing the data qualitatively as a backdrop or indirect guide. Therefore, the seismic analysis iterations could take multiple months with human expertise. The advancements in computer technology creates opportunities to develop automated tools for seismic interpretation that only a few years ago would have been prohibitively expensive. In this dissertation, I address the problem by investigating efficient seismic interpretation tools, designing related algorithms, and show the feasibility and effectiveness of applying them to various demanding interpretation problems on 2D/3D datasets. The tools are based on deep neural networks and employ convolutional layers to achieve artificial visual understanding of the datasets. First, I formulate salt detection as an image segmentation problem and develop a CNN to solve this problem with high efficiency and accuracy. CNNs with encoder-decoder architecture and skip-connections allows for extracting essential information from training data, thus results in high accuracy and great generalization across different type of datasets. Further extending from the segmentation end-to-end network framework, I introduce a recurrent style network for tracking irregular geobodies. The improvement is two-fold: the tracking algorithm allows for instance separation during segmentation, and the atomic design allows for more interaction on the user side to control the model application on various datasets. Apart from these supervised learning frameworks, I found that unsupervised learning provides even more powerful tools in other interpretation tasks. In the following chapter, I investigate the possibility to exploit the deep CNN architecture itself as a model parameterization method and perform image enhancing tasks. The deep network is optimized iteratively and can constrain the space of solutions to admissible models. Inspired by automatic recommendation system, in the next chapter, I propose a network that transforms seismic waveforms into a latent space in which they are aligned by similarities. Waveforms that belong to the same horizon, which are more similar to each other, can be extracted from the latent space more easily. In the last chapter, I propose a network architecture, plane-wave neural networks (PWNN), combining plane-wave destruction (PWD) filters and CNN into a single architecture. CNN can extract nonlinear features from spatial information, however, lacks the ability to understand spectral information. On the other hand, PWD filter, a local plane-wave model tailored specifically for representing seismic data, is effective to extract signals aligned along dominant seismic events. Finally, I discuss known limitations and suggest possible future research topics.Item Design and control of large collections of learning agents(2003) Agogino, Adrian Kujaneck; Ghosh, JoydeepThe intelligent control of multiple autonomous agents is an important yet difficult task. Previous methods used to address this problem have proved to be either too brittle, too hard to use, or not scalable to large systems. The Collective Intelligence project at NASA/Ames provides an elegant, machinelearning approach to address these problems. This approach mathematically defines some essential properties that a reward system should have to promote coordinated behavior among reinforcement learners. This thesis will focus on creating additional key properties and algorithms within the mathematics of the Framework of Collectives. The additions will allow agents to learn quickly in more complex systems. Also they will let agents learn with less knowledge of their environment. These additions will allow the framework to be applied more easily, to a much larger domain of multi-agent problems.Item Development of artificial intelligence methods for optical coherence tomography based medical diagnostic applications(2020-12-11) Baruah, Vikram Lal; Rylander, H. Grady (Henry Grady), 1948-; Milner, Thomas E.; Markey, Mia; Vargas, Gracie; Feldman, MarcPrimary challenges with early disease diagnosis are the lack of high-quality data and complexity of available data. Early diagnosis of both Atherosclerosis and AD, which effect hundreds of millions, is still limited by these two factors. Atherosclerosis is largely diagnosed decades after its onset with imaging lacking the resolution to directly identify critical disease components like vulnerable plaque with thin fibrous caps over a fibroatheroma (TCFA). Changes in the brain due to AD may begin 20 years before onset of disease symptoms and current diagnoses. Current diagnosis relies on low quality, late stage biomarkers like memory or motor function loss. In this work we have developed and used artificial intelligence methods to diagnose early onset of heart disease, cancer and Alzheimer’s with high sensitivity and specificity using Optical Coherence Tomography technology, surpassing the state of the art. Optical coherence tomography (OCT) can provide higher quality image data for Atherosclerosis and AD. OCT offers an axial resolution between 1-15 μm—one or two orders of magnitudes finer than conventional ultrasound, CT, and MRI (50-200 μm) while maintaining an imaging depth an order magnitude greater than microscopy. OCT can enable identification of key pathological features directly, like TCFA with a cap thickness of 65 μm. Despite these advantages, the complexity of OCT images makes interpretation difficult and volume of image data constrains real time analysis. Implementing Artificial Intelligence (AI) has the potential to make analysis of OCT images more accurate, consistent and rapid and reveal otherwise unseen features. AI can leverage machine learning and neural networks to learn optimal image features for each arterial tissue type. Furthermore, AI can be trained on ground truth guided by histology, the clinical gold standard, to identify commonly mistaken morphologies like TCFA and macrophages. While, various IVOCT plaque-classification approaches have been developed, they are poor medical diagnostics lacking in precision, reproducibility, and speed. In particular, the development of these approaches have been limited by a lack of comparison to histology (precision, reproducibility), or are not built using clinically standard IVOCT devices, require user-selected regions of interest (speed), or lack sensitivity to detect lipidic or calcified tissue (precision) (1,2,3,4). Expert human readers cannot reliably distinguish macrophages from TCFA and importantly are too slow for clinically relevant analysis of the hundreds of frames in an artery image. AI is well suited for medical diagnostics, as its training on gold standard data and optimized learning provides excellent precision. Lack of inter-operator variability with AI enhances reproducibility and GPU-powered AI provides a faster turn-around time compared to expert human readers, with near real-time results. In this dissertation I develop an automated histology validated plaque classification AI that identifies in IVOCT images three fundamental atherosclerotic tissue types: lipidic, fibrous, and calcific. Numerous studies (5) have demonstrated that neuropathology also presents in the retina, retinal OCT imaging is useful in characterizing and possibly providing early detection — the eye serving as a window to the brain. However, state-of-the-art retinal OCT analysis still focuses on macroscopic tissue features, like retinal nerve fiber layer thinning, which occur late in disease progression. A need is recognized to identify cellular biomarkers of neuropathology that occur early in AD. Work by Barsoum et al (5) have shown mitochondria as a promising early stage biomarker. Mitochondria are less tightly clustered and undergo fission during neurodegeneration. Light scattering theory (6) suggests that the angle distribution of back scattered light from mitochondria will differ between healthy and fission states. Thus, the dissertation evaluates whether a scattering angle resolved OCT (SAROCT) system can identify mitochondrial fission, a potential early stage AD biomarker. Studies by Gardner et al (6) have also demonstrated that SAROCT when combined with statistical Burr distribution fitting can identify retinal vascularity, another neuropathology biomarker. This dissertation extends this work by incorporating distribution fitting over a pixel panning window and convolutional neural networks to generate full angiography images from SAR OCT.Item Differences in the use of AI assistants : how human values influence AI assistant use or disuse(2018-05-03) Golden, Kathryn Elinor; Fleischmann, Kenneth R.This report is an analysis of the usage of artificial intelligence (AI) personal assistants such as Siri, Google Assistant, and Alexa through the examination of how an individual’s personal values influence their use of these devices. These assistants have become a built-in component of many technologies, and yet there is not a large amount of research on their utilization. Like most consumer level technologies, individual preferences determine how and when they will be used. Artificial assistants exist in a multitude of forms that most technology-using people will interact with, from bot assistance on websites or through the phone, to the personalized artificial intelligences used like the aforementioned Siri, Alexa, and Google Assistant. These specific assistants are utilized for everything from turning on the news to making purchases with the owner’s credit card information. They are privy to a multitude of personal information, and like most new technology, the level of comfort that people have using these devices varies depending on individual preferences. This report utilized a survey that focused on the Portrait Values Questionnaire created by Schwartz (2007) and made gender neutral by Verma, Fleischmann, and Koltai (2017) as well as in-depth, semi-structured, open-ended interviews. The ten interviews generated a greater understanding of individual perceptions of these devices and allowed for a more in depth look at specific examples and perspectives that strengthened the findings from the survey. The ultimate purpose of the report was to analyze how human values affect an individual’s use of these devices as one step towards a greater understanding of human values’ impact on technology, and how technology can be best created for humanity in turn.Item Discovering multi-purpose modules through deep multitask learning(2019-02-14) Meyerson, Elliot Keeler; Miikkulainen, Risto; Graumen, Kristen; Durrett, Greg; Nitschke, GeoffMachine learning scientists aim to discover techniques that can be applied across diverse sets of problems. Such techniques need to exploit regularities that are shared across tasks. This begs the question: What shared regularity is not yet being exploited? Complex tasks may share structure that is difficult for humans to discover. The goal of deep multitask learning is to discover and exploit this structure automatically by training a joint model across tasks. To this end, this dissertation introduces a deep multitask learning framework for collecting generic functional modules that are used in different ways to solve different problems. Within this framework, a progression of systems is developed based on assembling shared modules into task models and leveraging the complementary advantages of gradient descent and evolutionary optimization. In experiments, these systems confirm that modular sharing improves performance across a range of application areas, including general video game playing, computer vision, natural language processing, and genomics; yielding state-of-the-art results in several cases. The conclusion is that multi-purpose modules discovered by deep multitask learning can exceed those developed by humans in performance and generality.Item Engineering artificial intelligence systems for privacy(2023-08) Gohari, Parham; Topcu, Ufuk; Hale, Matthew; Klein, Adam; Vikalo, Haris; Chinchali, SandeepThis dissertation addresses the increasing need for privacy-aware artificial intelligence (AI) systems and investigates engineering approaches that strike a balance between maximizing performance and minimizing potential privacy threats. The privacy analysis methodology followed throughout this study uses the Contextual Integrity Theorem's framing of privacy as "appropriate information flows" to find a common vocabulary between the social concept of privacy and mathematical AI algorithms. The analysis begins with a technical privacy policy that describes the information flows that are appropriate for a given AI task. Once these permissible information flows are determined, the next step is to develop engineering approaches to optimize performance within the confinements of the technical privacy policy, which we refer to as privacy engineering. We study privacy engineering for AI systems in two sequential decision-making domains, namely policy synthesis and reinforcement learning. In both domains, agents must strategize their decisions to achieve an ultimate long-term goal, such as a robot navigating to a predetermined location. For policy synthesis, we study privacy engineering in the context of Markov decision processes (MDPs), which are abstract models of an environment consisting of states, actions, transition probabilities, and rewards. The privacy priority considered for this problem is to protect the confidentiality of the MDP's transition probabilities. Such a privacy priority is particularly relevant if the disclosure of the environment model to unauthorized parties could be harmful, such as the models that businesses develop to predict competitive market trends. For this problem, we identify differential privacy as an appropriate technical mechanism to achieve the set privacy goals and make two main contributions. First, we introduce the Dirichlet mechanism for enforcing differential privacy on simplex-valued data—which includes transition probabilities in MDPs with finite states and actions. Then, we use the Dirichlet mechanism to develop a differentially private policy synthesis algorithm. For the reinforcement learning problem, we study privacy engineering in the context of cooperative multi-agent reinforcement learning in which a team of agents must learn a common task through trial and error. For this problem, we assume that information disclosures about the agents' individual interactions with the environment violate privacy. We demonstrate that numerous existing algorithmic solutions rely on sharing environment interactions. Consequently, we introduce alternative privacy-engineered algorithms that establish permissible data-sharing frameworks according to the set technical privacy policy. The contributions of this dissertation demonstrate that privacy and AI can indeed be reconciled via privacy engineering. The findings highlight future research opportunities to design and implement AI algorithms with privacy as a priority.Item Evolutionary neural architecture search for deep learning(2019-02-08) Liang, Jason Zhi; Miikkulainen,, Risto; Stone, Peter; Baldick, Ross; Huang, QixingDeep neural networks (DNNs) have produced state-of-the-art results in many benchmarks and problem domains. However, the success of DNNs depends on the proper configuration of its architecture and hyperparameters. DNNs are often not used to their full potential because it is difficult to determine what architectures and hyperparameters should be used. While several approaches have been proposed, computational complexity of searching large design spaces makes them impractical for large modern DNNs. This dissertation introduces an efficient evolutionary algorithm (EA) for simultaneous optimization of DNN architecture and hyperparameters. It builds upon extensive past research of evolutionary optimization of neural network structure. Various improvements to the core algorithm are introduced, including: (1) discovering DNN architectures of arbitrary complexity; (1) generating modular, repetitive modules commonly seen in state-of-the-art DNNs; (3) extending to the multitask learning and multiobjective optimization domains; (4) maximizing performance and reducing wasted computation through asynchronous evaluations. Experimental results in image classification, image captioning, and multialphabet character recognition show that the approach is able to evolve networks that are competitive with or even exceed hand-designed networks. Thus, the method enables an automated and streamlined process to optimize DNN architectures for a given problem and can be widely applied to solve harder tasks.Item Evolving multimodal behavior through modular multiobjective neuroevolution(2014-05) Schrum, Jacob Benoid; Miikkulainen, RistoIntelligent organisms do not simply perform one task, but exhibit multiple distinct modes of behavior. For instance, humans can swim, climb, write, solve problems, and play sports. To be fully autonomous and robust, it would be advantageous for artificial agents, both in physical and virtual worlds, to exhibit a similar diversity of behaviors. This dissertation develops methods for discovering such behavior automatically using multiobjective neuroevolution. First, sensors are designed to allow multiple different interpretations of objects in the environment (such as predator or prey). Second, evolving networks are given ways of representing multiple policies explicitly via modular architectures. Third, the set of objectives is dynamically adjusted in order to lead the population towards the most promising areas of the search space. These methods are evaluated in five domains that provide examples of three different types of task divisions. Isolated tasks are separate from each other, but a single agent must solve each of them. Interleaved tasks are distinct, but switch back and forth within a single evaluation. Blended tasks do not have clear barriers, because an agent may have to perform multiple behaviors at the same time, or learn when to switch between opposing behaviors. The most challenging of the domains is Ms. Pac-Man, a popular classic arcade game with blended tasks. Methods for developing multimodal behavior are shown to achieve scores superior to other Ms. Pac-Man results previously published in the literature. These results demonstrate that complex multimodal behavior can be evolved automatically, resulting in robust and intelligent agents.
- «
- 1 (current)
- 2
- 3
- »