22484 research outputs found
Sort by
Multi-Scale Spatiotemporal Neural Computation: On the relationship between dynamical attractors, spiking neural networks, and convolutional neural circuits
No full textUnderstanding spatiotemporal neural dynamics and developing biologically-inspired artificial neural networks remain open challenges in computational neuroscience. Critical gaps persist in elucidating cortical rhythms, memory consolidation, and biological networks' remarkable spatiotemporal processing capabilities. This dissertation hypothesizes that asymmetric connectivity and dedicated fast-slow processing pathways in neural systems enhance depth, robustness, and versatility in handling complex spatiotemporal patterns. Our first contribution is elucidating how neurons communicate and synchronize activity via temporally precise spikes by examining the dynamics of spike-coding networks. Developing models of cortical neural oscillators reveal the origins of spontaneous transitions between active and silent states underlying slow-wave sleep rhythms, demonstrating how the intricate balance of excitation and inhibition orchestrates these oscillations. Our second is to establish a mathematical equivalence between Hopfield networks' associative memory models and spike-coding networks by showing that fast and slow asymmetric connectivity weights induce equivalent cyclic attractor dynamics in both systems. Introducing asymmetric weights in slow connections enables both models to learn and generate complex temporal firing sequences, transitioning between quasi-attractor states representing stored memories. Simulations demonstrate the efficacy of spike-coding networks for encoding and retrieving temporal sequences while performing the n-back working memory task. Our third contribution is to harness the potential of generative adversarial networks for unpaired cross-modality translation from 3 Tesla to 7 Tesla magnetic resonance imaging. We propose a fast-slow convolutional network architecture to enhance translation performance by balancing local and global information processing. This dissertation makes significant contributions by elucidating brain mechanisms underlying rhythms and memory, and unifying foundational computational frameworks while extracting principles to improve artificial neural network design
Engaging the Commuter Student: Examining the Impact of a Majors Mentor Program on Commuter Students
No full textUniversity commuter students are typically less engaged outside the classroom than their residential counterparts, being less likely to participate in extracurricular activities and university-sponsored events. This lack of engagement can lead to lower rates of persistence and retention. At the University of Pittsburgh at Greensburg, a small, regional, public university, commuter students comprise almost 60% of the overall student population. Despite being the majority, these students are typically less engaged on campus than the residential students and have reported being less satisfied in a number of areas. Using an improvement science approach, my theory of improvement was that by connecting engagement opportunities to academics, I could increase commuter students’ involvement. The intervention I tested was a Majors Mentor program which involved upper-level students mentoring and connecting with second semester first-year students within the same major or academic area. The impact of this peer mentoring program on commuter students’ engagement was measured through a mixed-methods approach, including attendance records, a post-participation satisfaction and opinion survey, and semi-structured interviews conducted with the commuter students who were both mentees and mentors. Results of this study suggest that the Majors Mentor program helped commuter students to connect more actively with the campus community, and that they benefitted from the intervention, engaging more with the institution. Strengths, limitations, implications, and suggested improvements to the intervention are also discussed. Further iterations of this intervention are recommended in order to collect more conclusive data and findings and to help improve the program in the future
"Exploring Environmental Factors in Neurosurgical and Ophthalmology Operating Rooms to Mitigate Surgical Site Infections: An Observation Study."
No full textBackground: Surgical site infections (SSIs) are infections that patients can acquire after undergoing surgery in a hospital. These infections are quite common and can pose a significant risk to the patient's health. The operating room (OR) is a sterile space with regulated airflow, humidity, and pressure to ensure a clean environment. OR traffic should be minimized to maintain sterility and reduce SSI risks.
Methods: This observational study aimed to understand how operating room staff behavior and environmental factors affect surgical outcomes. The study was conducted for 4 weeks at a teaching hospital and focused on 10 neurosurgical procedures. The research student leading the study observed OR traffic patterns and documented environmental parameters such as temperature, humidity, and pressure. Automated data acquisition was implemented for ophthalmology cases.
Results: The study found correlations between particulate sizes and room conditions in an operating room. 0.3μm particulate size had a moderate to strong positive correlation of 0.737, while 1.0μm showed a very weak positive correlation of 0.087. The 5.0μm particulate size had a moderate positive correlation of 0.344, with 11.8% variability attributed to observed operating room traffic. The study also noted mild fluctuations in temperature, humidity, and pressure within the operating room.
Conclusion: A prospective research study conducted at a university-affiliated teaching hospital suggests a potential link between the OR environment and the risk of developing surgical site infections (SSIs). However, the study's limitations and small sample size must be considered when interpreting the findings. Future research should address these constraints for accurate results. Understanding the OR environment is crucial in preventing SSIs, improving patient outcomes, and reducing the burden of postoperative complications.
Public Health Significance: Surgical site infections (SSIs) are of a significant concern in the healthcare industry, particularly in neurosurgery and ophthalmology. SSIs result in physical discomfort, extended hospital stays, increased healthcare costs, and mortality. By prioritizing prevention of SSIs, healthcare resources can be utilized more efficiently, reducing the economic burden on the healthcare system. Controlling SSIs is crucial for maintaining high standards of patient care, reducing healthcare burden, preventing infections, and fostering a more equitable healthcare system
Effects of orthographic similarity, individual differences, and training manipulation on learning translation-ambiguous and unambiguous words
No full textLearning a second language (L2) as an adult is a challenging endeavor, requiring the integration of the L2 into an established first language (L1) system. Throughout their linguistic journey, L2 learners will also encounter L2 vocabulary that has two or more translations across languages (so-called translation-ambiguous words). These words are particularly difficult to learn (Degani & Tokowicz, 2010) and to process (Laxén & Lavaur, 2010). Moreover, even highly proficient bilinguals have difficulty processing translation-ambiguous cognate words (Boada et al., 2013). In three experiments, we investigated adult L2 learning of translation-ambiguous words with varying orthographic similarity between translations (i.e., ranging from noncognates to cognates with partial overlap to cognates with complete overlap). For instance, the Dutch-English word pair "jongen-boy" represents lower orthographic similarity, whereas the Dutch-English word pair "auteur-author" represents higher orthographic similarity, and the Dutch-English word pair "fruit-fruit" represents the highest level of orthographic similarity. In Experiment 1, we investigated whether the translation-ambiguity disadvantage on bilinguals’ cognate processing (Boada et al.) is also present in adult L2 learning of words with varying orthographic similarity. Results revealed that during the beginning stages of vocabulary learning, the translation-ambiguity disadvantage was present for words with higher levels of orthographic similarity but surprisingly not for words with lower orthographic similarity. In Experiment 2, we explored the association between individual differences and L2 learning of translation-ambiguous and unambiguous words with varying orthographic similarity. Results showed that lower working memory (WM) individuals showed no translation-ambiguity disadvantage, while higher WM individuals experienced it for words with moderate to high orthographic similarity. Building on these findings, Experiment 3 explored whether a training manipulation could enhance the learning of translation-ambiguous words with varying orthographic similarity. Results uncovered a complex interaction between the training manipulation, the test, and either phonological short-term memory (PSTM) or WM. Intriguingly, the impact of the training manipulation varies across orthographic similarity levels, significantly affecting lower levels. These findings provide nuanced perspectives on L2 vocabulary learning, offering valuable insights into how L2 learners process and represent translation-ambiguous words. Implications for models of translation ambiguity are discussed, emphasizing the multifaceted nature of L2 vocabulary learning
Robot Locomotion through Tunable Spiking and Bursting Rhythms using Efficient Bio-mimetic Neural Networks on Loihi and Arduino Platforms
No full textRhythmic tasks that biological beings perform such as breathing, walking, and swimming, use specialized neural networks called central pattern generators (CPG). This paper aims to take this concept further by designing and implementing a tunable bursting central pattern generator to control quadruped robots for the first time, to the best of our knowledge. Bursting CPGs allow for more granular control over the motion and speed of operation while retaining the low memory usage and latency capabilities of spiking CPGs. A bio-mimetic neuron model is chosen for this implementation which is highly optimized to run real-time on standard (Arduino microcontroller) and specialized (Intel Loihi) hardware. The Petoi bittle is chosen as the model hardware setup to showcase the efficiency of the proposed CPGs even in serial processing architectures. The CPG network is also realized in a completely asynchronous Loihi architecture to illustrate its versatility. The fully connected network running on CPG takes around 10 kilo bytes of memory (33% of Arduino capacity) to execute different modes of locomotion - walk, jump, trot, gallop, and crawl. Benchmarking results show that the bio-mimetic neurons take around 600 bytes (around 2%) more memory than Izhikevich neurons while being 0.02ms (around 14%) faster in isolated neuron testing
Polymer Infiltration and Pyrolysis of Silicon Carbide Ceramics Using Transient Liquid
No full textThe polymer infiltration and pyrolysis (PIP) method has a limitation in increasing the density of SiC ceramics after multiple PIP cycles. This is due to an increasing number of blocked pores that decreased the amount of polymer infiltrated deep into the interior of bulk materials. In this study, a new process that incorporates Ni and carbon nanoparticles into SiC polymer precursor is examined. Ni and carbon nanoparticles were uniformly distributed into porous SiC ceramics during the polymer infiltration. The reaction of nanomaterials with SiC polymer precursor results in a transient liquid phase during the pyrolysis, which mitigates the pore closure. Ni nanoparticles reacted with the SiC precursor to form a nickel silicide of low melting temperature such as Ni2Si and NiSi phases. During high temperature pyrolysis, these silicide phases turned to a liquid phase, facilitated the redistribution of the infiltrated material, and maintained the pore structure open. In later infiltration steps, the co-addition of carbon nanoparticles into the polymer precursor helps the conversion of nickel silicide to nickel carbide and decreases the amount of residual nickel silicide which may be harmful for mechanical strength at high temperature. Results of this study show significant improvement in the density of SiC ceramics in comparison to traditional PIP
Coherent Photonic Vector Processor for Scalable and Efficient Optical Computing.
No full textRecent advancements in artificial intelligence (AI) have hinged on the capacity to train increasingly vast parameter sets within neural networks, a trend that has outpaced the computational capabilities of traditional digital hardware. This has led to a critical need for new computing paradigms that can simultaneously enhance computational efficiency and throughput. Photonic analog computing emerges as a promising solution to reduce latency and boost performance, particularly in addressing the memory-related challenges that underpin AI's success. This thesis presents an overview of the integration of optical and electrical memories with photonic hardware, underlining the pivotal role of memory technologies. It emphasizes the need for more efficient and compact memory for optical computing and presents an integrated coherent solution to process temporally multiplexed optical signals using a modular dot-product unit cell. We use these unit cells to demonstrate multiply accumulate operations on real- and complex-valued inputs using coherent detection and temporal integration. We then extend this to computing the covariance between stochastic bit streams which can be used to estimate correlation between data streams in the optical domain. Finally, we demonstrate a path to scaling up our platform to enable general matrix-matrix operations. This approach has the potential to enable highly efficient and scalable optical computing on-chip for a broad variety of AI applications
Paving the Way for Protein (Un)Folding Pathways: Towards Machine-Learning Guided Weighted Ensemble Simulations of Rare-Event Sampling
No full textMolecular dynamics simulations of rare events such as protein folding and unfolding are difficult due the large differences in time scales between the process of interest and stable states. In my dissertation, I will present advances to the weighted ensemble (WE) path sampling strategy that I have developed and demonstrate how the WE strategy can be used to efficiently simulate a protein-folding process. In Chapter 1, I motivate the need for studying biological processes using molecular dynamics simulations, introduce the WE method, and present potential future directions. In Chapters 2 and 3, I will detail joint experimental-simulation studies in which I have characterized both the thermodynamics and kinetics of a folding process for a small alpha-helical protein and five artificially modified variants of this protein. In the second half of my thesis, I will further describe some of the improvements to the WE strategy that allow for simulation of long time scale processes. In Chapter 4, I will detail changes in the open-source WESTPA software package for running WE simulations. In Chapter 5, I will present how machine learning-based progress coordinates can be used to accelerate WE simulation sampling on-the-fly. These chapters, together, demonstrate the current capabilities of the WE strategy, including enhancements provided by machine learning
Invisible Protein States and How to View Them: Integrating 19F NMR and Weighted Ensemble Simulations
No full textProtein conformational dynamics are a keystone to understanding biology at the mechanistic level. Traditional methods for studying conformational dynamics are to use NMR spectroscopy and molecular dynamics (MD) simulations. However, many biologically interesting events occur outside of the timescales accessible to conventional MD simulations, and using NMR, many transient states can be sparsely populated, or ensemble averaged, making their resolution recalcitrant to traditional NMR approaches. To resolve these dynamic protein states, both enhanced simulation methods and highly sensitive 19F NMR experiments are needed. In this dissertation, I integrate and further develop computational and experimental methods to fully characterize the conformational ensemble of the HIV-1 capsid protein CTD dimer. In Chapter 1, I motivate the need for a combination of weighted ensemble (WE) path sampling simulations and 19F NMR to resolve the multi-state conformational dynamics of the HIV-1 capsid protein. Chapter 2 describes the development and validation of new force field parameters that allow for simulations of the fluorinated amino acids commonly used in 19F NMR experiments. In Chapter 3, I present a comprehensive study of the structures and dynamics of an "invisible" alternate state of the HIV-1 capsid protein CTD dimer using WE simulations and 19F NMR experiments. A bottleneck for this project was the identification of effective parameters for WE simulations. In Chapter 4, I developed a new algorithm for WE resampling which automates tedious parts of WE simulation parameter selection using concepts from reinforcement learning. Finally, in Chapter 5, I developed a software package for convenient WE data analysis and plotting. Together, the above chapters demonstrate the power of integrating simulations and experiments, as well as the potential of more robust path sampling methods and software for characterizing the conformational ensembles of proteins
Common Read Discussion of "The Data Detective: Ten Easy Rules to Make Sense of Statistics" by Tim Harford
No full textPowerPoint presentation and supplementary materials to facilitate the book discussion and group activities for the "The Data Detective: Ten Easy Rules to Make Sense of Statistics" by Tim Harford