Search

AmberTools is a free and open-source collection of programs used to set up, run, and analyze molecular simulations. The newer features contained within AmberTools23 are briefly described in this Application note.

Current deep-learning models for object recognition are known to be heavily biased toward texture. In contrast, human visual systems are known to be biased toward shape and structure. What could be the design principles in human visual systems that led to this difference? How could we introduce more shape bias into the deep learning models? In this paper, we report that sparse coding, a ubiquitous principle in the brain, can in itself introduce shape bias into the network. We found that enforcing the sparse coding constraint using a non-differential Top-K operation can lead to the emergence of structural encoding in neurons in convolutional neural networks, resulting in a smooth decomposition of objects into parts and subparts and endowing the networks with shape bias. We demonstrated this emergence of shape bias and its functional benefits for different network structures with various datasets. For object recognition convolutional neural networks, the shape bias leads to greater robustness against style and pattern change distraction. For the image synthesis generative adversary networks, the emerged shape bias leads to more coherent and decomposable structures in the synthesized images. Ablation studies suggest that sparse codes tend to encode structures, whereas the more distributed codes tend to favor texture. Our code is host at the github repository: \url{https://github.com/Crazy-Jack/nips2023_shape_vs_texture}, Comment: Published as NeurIPS 2023 (Oral)

Deep learning models are known to exhibit a strong texture bias, while human tends to rely heavily on global shape structure for object recognition. The current benchmark for evaluating a model's global shape bias is a set of style-transferred images with the assumption that resistance to the attack of style transfer is related to the development of global structure sensitivity in the model. In this work, we show that networks trained with style-transfer images indeed learn to ignore style, but its shape bias arises primarily from local detail. We provide a \textbf{Disrupted Structure Testbench (DiST)} as a direct measurement of global structure sensitivity. Our test includes 2400 original images from ImageNet-1K, each of which is accompanied by two images with the global shapes of the original image disrupted while preserving its texture via the texture synthesis program. We found that \textcolor{black}{(1) models that performed well on the previous cue-conflict dataset do not fare well in the proposed DiST; (2) the supervised trained Vision Transformer (ViT) lose its global spatial information from positional embedding, leading to no significant advantages over Convolutional Neural Networks (CNNs) on DiST. While self-supervised learning methods, especially mask autoencoder significantly improves the global structure sensitivity of ViT. (3) Improving the global structure sensitivity is orthogonal to resistance to style-transfer, indicating that the relationship between global shape structure and local texture detail is not an either/or relationship. Training with DiST images and style-transferred images are complementary, and can be combined to train network together to enhance the global shape sensitivity and robustness of local features.} Our code will be hosted in github: https://github.com/leelabcnbc/DiST

The visual system evolved to process natural scenes, yet most of our understanding of the topology and function of visual cortex derives from studies using artificial stimuli. To gain deeper insights into visual processing of natural scenes, we utilized widefield calcium-imaging of primate V4 in response to many natural images, generating a large dataset of columnar-scale responses. We used this dataset to build a digital twin of V4 via deep learning, generating a detailed topographical map of natural image preferences at each cortical position. The map revealed clustered functional domains for specific classes of natural image features. These ranged from surface-related attributes like color and texture to shape-related features such as edges, curvature, and facial features. We validated the model-predicted domains with additional widefield calcium-imaging and single-cell resolution two-photon imaging. Our study illuminates the detailed topological organization and neural codes in V4 that represent natural scenes., Comment: 39 pages, 14 figures

Dark Forest by Tai Wen Lee Dark Forest is an interactive game where you play as someone that woke up in an unfamiliar location. You are presented with choices to traverse the environment you're in as you figure out what is going on.

We explore the use of a neural network inspired by predictive coding for modeling human music perception. This network was developed based on the computational neuroscience theory of recurrent interactions in the hierarchical visual cortex. When trained with video data using self-supervised learning, the model manifests behaviors consistent with human visual illusions. Here, we adapt this network to model the hierarchical auditory system and investigate whether it will make similar choices to humans regarding the musicality of a set of random pitch sequences. When the model is trained with a large corpus of instrumental classical music and popular melodies rendered as mel spectrograms, it exhibits greater prediction errors for random pitch sequences that are rated less musical by human subjects. We found that the prediction error depends on the amount of information regarding the subsequent note, the pitch interval, and the temporal context. Our findings suggest that predictability is correlated with human perception of musicality and that a predictive coding neural network trained on music can be used to characterize the features and motifs contributing to human perception of music., Comment: 5 pages, 5 figures, currently in peer review

Recent advances in deep generative models have led to immense progress in 3D shape synthesis. While existing models are able to synthesize shapes represented as voxels, point-clouds, or implicit functions, these methods only indirectly enforce the plausibility of the final 3D shape surface. Here we present a 3D shape synthesis framework (SurfGen) that directly applies adversarial training to the object surface. Our approach uses a differentiable spherical projection layer to capture and represent the explicit zero isosurface of an implicit 3D generator as functions defined on the unit sphere. By processing the spherical representation of 3D object surfaces with a spherical CNN in an adversarial setting, our generator can better learn the statistics of natural shape surfaces. We evaluate our model on large-scale shape datasets, and demonstrate that the end-to-end trained model is capable of generating high fidelity 3D shapes with diverse topology., Comment: ICCV 2021. Project page: https://github.com/aluo-x/NeuralRaycaster

Feedforward CNN models have proven themselves in recent years as state-of-the-art models for predicting single-neuron responses to natural images in early visual cortical neurons. In this paper, we extend these models with recurrent convolutional layers, reflecting the well-known massive recurrence in the cortex, and show robust increases in predictive performance over feedforward models across thousands of hyperparameter combinations in three datasets of macaque V1 and V2 single-neuron responses. We propose the recurrent circuit can be conceptualized as a form of ensemble computing, with each iteration generating more effective feedforward paths of various path lengths to allow a combination of solutions in the final approximation. The statistics of the paths in the ensemble provide insights to the differential performance increases among our recurrent models. We also assess whether the recurrent circuits learned for neural response prediction can be related to cortical circuits. We find that the hidden units in the recurrent circuits of the appropriate models, when trained on long-duration wide-field image presentations, exhibit similar temporal response dynamics and classical contextual modulations as observed in V1 neurons. This work provides insights to the computational rationale of recurrent circuits and suggests that neural response prediction could be useful for characterizing the recurrent neural circuits in the visual cortex.

© 2019 The Royal Society of Chemistry. Bis(trifluoromethylsulfonyl)imide (TFSI)-based ionic liquid (IL) has high thermal and electrochemical stability, but it is not an ideal battery electrolyte due to the poor rate capability of cells that use it, problematic anode compatibility, and high cost. The incorporation of a carbonate solvent could mitigate these problems, but it would also lead to serious Al current collector corrosion at high potential. This long-existing problem is overcome in this study by modulating the LiTFSI concentration and IL/carbonate ratio in the hybrid electrolyte. The Al corrosion and electrolyte decomposition side reactions at 5 V (vs. Li+/Li) can be suppressed in 3 M LiTFSI 25%-IL electrolyte, in which good performance of a high-voltage LiNi0.5Mn1.5O4 (LNMO) cathode is achieved. Capacities of 140 and 88 mA h g-1 were measured at 0.1 and 2C, respectively (vs. 25 mA h g-1 at 2C for a plain LiTFSI/PMP-TFSI IL electrolyte). After 300 charge-discharge cycles, 90% of the initial LNMO capacity was retained. This electrolyte also shows low flammability and great wettability toward a polyethylene separator. Moreover, this electrolyte allows elevatederature storage and operation of LNMO cells at 55 °C, which is not possible with the conventional carbonate electrolyte. Good compatibility of the electrolyte with a graphite anode is also demonstrated. The proposed electrolyte design concept has great potential for next-generation 5 V Li-ion batteries.

Our brain perceives the world by exploiting multisensory cues to extract information about various aspects of external stimuli. The sensory cues from the same stimulus should be integrated to improve perception, and otherwise segregated to distinguish different stimuli. In reality, however, the brain faces the challenge of recognizing stimuli without knowing in advance the sources of sensory cues. To address this challenge, we propose that the brain conducts integration and segregation concurrently with complementary neurons. Studying the inference of heading-direction via visual and vestibular cues, we develop a network model with two reciprocally connected modules modeling interacting visual-vestibular areas. In each module, there are two groups of neurons whose tunings under each sensory cue are either congruent or opposite. We show that congruent neurons implement integration, while opposite neurons compute cue disparity information for segregation, and the interplay between two groups of neurons achieves efficient multisensory information processing.

Chia-Jung Liu,1 Tai-Fen Lee,2 Sheng-Yuan Ruan,1 Chong-Jen Yu,1 Jung-Yien Chien,1 Po-Ren Hsueh1,21Department of Internal Medicine, National Taiwan University Hospital, National Taiwan University College of Medicine, Taipei, Taiwan; 2Department of Laboratory Medicine, National Taiwan University Hospital, National Taiwan University College of Medicine, Taipei, TaiwanObjectives: The incidence of Pneumocystis pneumonia (PCP) has been increasing among non-HIV-infected patients. Here, we investigated the clinical characteristics, treatment outcomes, and prognostic factors of PCP in non-HIV-infected patients.Patients and methods: Information on clinical characteristics, treatment outcomes, and prognostic factors of PCP patients who were treated at a medical center in northern Taiwan from October 2015 to October 2016 were retrieved from medical records and evaluated.Results: Among the patients with PCP included in the study, 84 were non-HIV-infected and 25 were HIV-infected. Non-HIV-infected patients with PCP had a longer duration between radiographic findings and treatment (P<0.001), and a higher rate of hospital-associated PCP (P<0.001), hypoxia (P=0.015), respiratory failure (P<0.001), and mortality (P=0.006) than HIV-infected patients with PCP. Among non-HIV-infected patients, non-survivors had a higher fungal burden (46.2% vs 22.2%, P=0.039), higher requirement for adjunctive steroid treatment (94.9% vs 71.1%, P=0.011), and higher rate of pneumothorax (17.9% vs 2.2%, P=0.038) than survivors. Multiple logistic regression revealed that lymphopenia (odds ratio [OR] =3.24, 95% confidence interval [CI] =1.07–9.79; P=0.037), adjunctive steroid use (OR =6.23, 95% CI =1.17–33.14; P=0.032), and pneumothorax (OR =10.68, 95% CI =1.00–113.93; P=0.050) were significantly associated with increased 60-day mortality among non-HIV-infected PCP patients.Conclusion: Lymphopenia, adjunctive steroid therapy, and pneumothorax were significan

The abundant recurrent horizontal and feedback connections in the primate visual cortex are thought to play an important role in bringing global and semantic contextual information to early visual areas during perceptual inference, helping to resolve local ambiguity and fill in missing details. In this study, we find that introducing feedback loops and horizontal recurrent connections to a deep convolution neural network (VGG16) allows the network to become more robust against noise and occlusion during inference, even in the initial feedforward pass. This suggests that recurrent feedback and contextual modulation transform the feedforward representations of the network in a meaningful and interesting way. We study the population codes of neurons in the network, before and after learning with feedback, and find that learning with feedback yielded an increase in discriminability (measured by d-prime) between the different object classes in the population codes of the neurons in the feedforward path, even at the earliest layer that receives feedback. We find that recurrent feedback, by injecting top-down semantic meaning to the population activities, helps the network learn better feedforward paths to robustly map noisy image patches to the latent representations corresponding to important visual concepts of each object class, resulting in greater robustness of the network against noises and occlusion as well as better fine-grained recognition., Comment: 10 pages, 5 figures

In this paper we developed a hierarchical network model, called Hierarchical Prediction Network (HPNet), to understand how spatiotemporal memories might be learned and encoded in the recurrent circuits in the visual cortical hierarchy for predicting future video frames. This neurally inspired model operates in the analysis-by-synthesis framework. It contains a feed-forward path that computes and encodes spatiotemporal features of successive complexity and a feedback path for the successive levels to project their interpretations to the level below. Within each level, the feed-forward path and the feedback path intersect in a recurrent gated circuit, instantiated in a LSTM module, to generate a prediction or explanation of the incoming signals. The network learns its internal model of the world by minimizing the errors of its prediction of the incoming signals at each level of the hierarchy. We found that hierarchical interaction in the network increases semantic clustering of global movement patterns in the population codes of the units along the hierarchy, even in the earliest module. This facilitates the learning of relationships among movement patterns, yielding state-of-the-art performance in long range video sequence predictions in the benchmark datasets. The network model automatically reproduces a variety of prediction suppression and familiarity suppression neurophysiological phenomena observed in the visual cortex, suggesting that hierarchical prediction might indeed be an important principle for representational learning in the visual cortex., Comment: Some of the results are not replicable

Recurrent feedback connections in the mammalian visual system have been hypothesized to play a role in synthesizing input in the theoretical framework of analysis by synthesis. The comparison of internally synthesized representation with that of the input provides a validation mechanism during perceptual inference and learning. Inspired by these ideas, we proposed that the synthesis machinery can compose new, unobserved images by imagination to train the network itself so as to increase the robustness of the system in novel scenarios. As a proof of concept, we investigated whether images composed by imagination could help an object recognition system to deal with occlusion, which is challenging for the current state-of-the-art deep convolutional neural networks. We fine-tuned a network on images containing objects in various occlusion scenarios, that are imagined or self-generated through a deep generator network. Trained on imagined occluded scenarios under the object persistence constraint, our network discovered more subtle and localized image features that were neglected by the original network for object classification, obtaining better separability of different object classes in the feature space. This leads to significant improvement of object recognition under occlusion for our network relative to the original network trained only on un-occluded images. In addition to providing practical benefits in object recognition under occlusion, this work demonstrates the use of self-generated composition of visual scenes through the synthesis loop, combined with the object persistence constraint, can provide opportunities for neural networks to discover new relevant patterns in the data, and become more flexible in dealing with novel situations., Comment: Accepted by 14th Conference on Computer and Robot Vision

We develop an approach for unsupervised learning of associations between co-occurring perceptual events using a large graph. We applied this approach to successfully solve the image captcha of China's railroad system. The approach is based on the principle of suspicious coincidence. In this particular problem, a user is presented with a deformed picture of a Chinese phrase and eight low-resolution images. They must quickly select the relevant images in order to purchase their train tickets. This problem presents several challenges: (1) the teaching labels for both the Chinese phrases and the images were not available for supervised learning, (2) no pre-trained deep convolutional neural networks are available for recognizing these Chinese phrases or the presented images, and (3) each captcha must be solved within a few seconds. We collected 2.6 million captchas, with 2.6 million deformed Chinese phrases and over 21 million images. From these data, we constructed an association graph, composed of over 6 million vertices, and linked these vertices based on co-occurrence information and feature similarity between pairs of images. We then trained a deep convolutional neural network to learn a projection of the Chinese phrases onto a 230-dimensional latent space. Using label propagation, we computed the likelihood of each of the eight images conditioned on the latent space projection of the deformed phrase for each captcha. The resulting system solved captchas with 77% accuracy in 2 seconds on average. Our work, in answering this practical challenge, illustrates the power of this class of unsupervised association learning techniques, which may be related to the brain's general strategy for associating language stimuli with visual objects on the principle of suspicious coincidence., Comment: 8 pages, 7 figures, 14th Conference on Computer and Robot Vision 2017

We develop a model of perceptual similarity judgment based on re-training a deep convolution neural network (DCNN) that learns to associate different views of each 3D object to capture the notion of object persistence and continuity in our visual experience. The re-training process effectively performs distance metric learning under the object persistency constraints, to modify the view-manifold of object representations. It reduces the effective distance between the representations of different views of the same object without compromising the distance between those of the views of different objects, resulting in the untangling of the view-manifolds between individual objects within the same category and across categories. This untangling enables the model to discriminate and recognize objects within the same category, independent of viewpoints. We found that this ability is not limited to the trained objects, but transfers to novel objects in both trained and untrained categories, as well as to a variety of completely novel artificial synthetic objects. This transfer in learning suggests the modification of distance metrics in view- manifolds is more general and abstract, likely at the levels of parts, and independent of the specific objects or categories experienced during training. Interestingly, the resulting transformation of feature representation in the deep networks is found to significantly better match human perceptual similarity judgment than AlexNet, suggesting that object persistence could be an important constraint in the development of perceptual similarity judgment in biological neural networks., Comment: Accepted to ICLR2017

The relationship between formation of active in-line attack conformations and monovalent (Na(+)) and divalent (Mg(2+)) metal ion binding in hammerhead ribozyme (HHR) has been explored with molecular dynamics simulations. To stabilize repulsions between negatively charged groups, different requirements of the threshold occupancy of metal ions were observed in the reactant and activated precursor states both in the presence and in the absence of a Mg(2+) in the active site. Specific bridging coordination patterns of the ions are correlated with the formation of active in-line attack conformations and can be accommodated in both cases. Furthermore, simulation results suggest that the HHR folds to form an electronegative recruiting pocket that attracts high local concentrations of positive charge. The present simulations help to reconcile experiments that probe the metal ion sensitivity of HHR catalysis and support the supposition that Mg(2+), in addition to stabilizing active conformations, plays a specific chemical role in catalysis.

Replica Exchange (RE) simulations have emerged as an important algorithmic tool for the molecular sciences. RE simulations involve the concurrent execution of independent simulations which infrequently interact and exchange information. The next set of simulation parameters are based upon the outcome of the exchanges. Typically RE functionality is integrated into the molecular simulation software package. A primary motivation of the tight integration of RE functionality with simulation codes has been performance. This is limiting at multiple levels. First, advances in the RE methodology are tied to the molecular simulation code. Consequently these advances remain confined to the molecular simulation code for which they were developed. Second, it is difficult to extend or experiment with novel RE algorithms, since expertise in the molecular simulation code is typically required. In this paper, we propose the RepEx framework which address these aforementioned shortcomings of existing approaches, while striking the balance between flexibility (any RE scheme) and scalability (tens of thousands of replicas) over a diverse range of platforms. RepEx is designed to use a pilot-job based runtime system and support diverse RE Patterns and Execution Modes. RE Patterns are concerned with synchronization mechanisms in RE simulation, and Execution Modes with spatial and temporal mapping of workload to the CPU cores. We discuss how the design and implementation yield the following primary contributions of the RepEx framework: (i) its ability to support different RE schemes independent of molecular simulation codes, (ii) provide the ability to execute different exchange schemes and replica counts independent of the specific availability of resources, (iii) provide a runtime system that has first-class support for task-level parallelism, and (iv) required scalability along multiple dimensions., Comment: 12 pages, 13 figures

Molecular dynamics simulations have been performed to investigate the role of Mg2+ in the full-length hammerhead ribozyme cleavage reaction. In particular, the aim of this work is to characterize the binding mode and conformational events that give rise to catalytically active conformations and stabilization of the transition state. Toward this end, a series of eight 12 ns molecular dynamics simulations have been performed with different divalent metal binding occupations for the reactant, early and late transition state using recently developed force field parameters for metal ions and reactive intermediates in RNA catalysis. In addition, hybrid QM/MM calculations of the early and late transition state were performed to study the proton-transfer step in general acid catalysis that is facilitated by the catalytic Mg2+ ion. The simulations suggest that Mg2+ is profoundly involved in the hammerhead ribozyme mechanism both at structural and catalytic levels. Binding of Mg2+ in the active site plays a key structural role in the stabilization of stem I and II and to facilitate formation of near attack conformations and interactions between the nucleophile and G12, the implicated general base catalyst. In the transition state, Mg2+ binds in a bridging position where it stabilizes the accumulated charge of the leaving group while interacting with the 2'OH of G8, the implicated general acid catalyst. The QM/MM simulations provide support that, in the late transition state, the 2'OH of G8 can transfer a proton to the leaving group while directly coordinating the bridging Mg2+ ion. The present study provides evidence for the role of Mg2+ in hammerhead ribozyme catalysis. The proposed simulation model reconciles the interpretation of available experimental structural and biochemical data, and provides a starting point for more detailed investigation of the chemical reaction path with combined QM/MM methods.

Results of a series of 12 ns molecular dynamics (MD) simulations of the reactant state (with and without a Mg(2+) ion), early and late transition state mimics are presented based on a recently reported crystal structure of a full-length hammerhead RNA. The simulation results support a catalytically active conformation with a Mg(2+) ion bridging the A9 and scissile phosphates. In the reactant state, the Mg(2+) spends significant time closely associated with the 2'OH of G8, but remains fairly distant from the leaving group O(5') position. In the early TS mimic simulation, where the nucleophilic O(2') and leaving group O(5') are equidistant from the phosphorus, the Mg(2+) ion remains tightly coordinated to the 2'OH of G8, but is positioned closer to the O(5') leaving group, stabilizing the accumulating charge. In the late TS mimic simulation, the coordination around the bridging Mg(2+) ion undergoes a transition whereby the coordination with the 2'OH of G8 is replace by the leaving group O(5') that has developed significant charge. At the same time, the 2'OH of G8 forms a hydrogen bond with the leaving group O(5') and is positioned to act as a general acid catalyst. This work represents the first reported simulations of the full-length hammerhead structure and TS mimics, and provides direct evidence for the possible role of a bridging Mg(2+) ion in catalysis that is consistent with both crystallographic and biochemical data.

In this study, first, we prepared ternary component semiconductor powders of ZnmIn2S3+m (ZIS) by using microwave-assisted hydrothermal method. Microwaveassisted hydrothermal process is a facile way to produce nanomaterials at a shorter reaction time with less secondary impurities while changing the amount of zinc in a series of solid solutions. With little adjustment of the ratios of [Zn]/[In], the hydrogen production rate of the photocatalysts, particle sizes, and band gap are significantly different. In the following experiments, the core-shell of nanoshell@SiO2, as well as the nanostructure of photocatalyst, were further investigated. Solar energy in the visible-light range is expected to be absorbed by the photocatalyst first without any interference from the metal nanoshells. The presence of metal nanoshells as the core can absorb the solar energy in the IR and visible-light region ranging from 500 nm to 900 nm. Our data showed that the plasmonic-enhanced photocatalytic activity was a function of the absorption of Au nanoshells. At the absorption wavelength of 500 nm of the Au nanoshells, the enhancement of hydrogen production was probably due to the plasmonic effect. As opposed to the absorption wavelength at 900 nm, the enhancement was due to thermal effect. Furthermore, the nanoshell absorbing at 700 nm has the highest hydrogen production rate which may be due to both the plasmonic and thermal effect. Thickness of the silica layer also influenced hydrogen evolution. Steady-state photoluminescence measurement was used to investigate the electron-hole recombination mechanism. Results demonstrated that efficient electron-hole charge separation can be achieved when Au nanoshells were incorporated. Finally, the core-shell with various [Zn]/[In] compositions was used to identify the interaction between the Au nanoshell and the photocatalyst.

Chang-Moon Lee,1 Tai Kyoung Lee,2–5 Dae-Ik Kim,1,6 Yu-Ri Kim,7 Meyoung-Kon Kim,7 Hwan-Jeong Jeong,2–5 Myung-Hee Sohn,2–5 Seok Tae Lim2–5 1Department of Biomedical Engineering, Chonnam National University, Yeosu, Jeollanam-Do, Republic of Korea; 2Department of Nuclear Medicine, Chonbuk National University Medical School and Hospital, Jeonju, Jeollabuk-Do, Republic of Korea; 3Cyclotron Research Center, Chonbuk National University Medical School and Hospital, Jeonju, Jeollabuk-Do, Republic of Korea; 4Biomedical Research Institute, Chonbuk National University Medical School and Hospital, Jeonju, Jeollabuk-Do, Republic of Korea; 5Molecular Imaging and Therapeutic Medicine Research Center, Chonbuk National University Medical School and Hospital, Jeonju, Jeollabuk-Do, Republic of Korea; 6School of Electrical, Electronic Communication, and Computer Engineering, Chonnam National University, Yeosu, Jeollanam-Do, Republic of Korea; 7Department of Biochemistry and Molecular Biology, Korea University Medical School and College, Seounbuk-Gu, Seoul, Republic of Korea Purpose: In this study, we investigated the absorption and distribution of rhodamine B isothiocyanate (RITC)-incorporated silica oxide nanoparticles(SiNPs) (RITC-SiNPs) after oral exposure, by conducting optical imaging, with a focus on tracking the movement of RITC-SiNPs of different particle size and surface charge. Methods: RITC-SiNPs (20 or 100 nm; positively or negatively charged) were used to avoid the dissociation of a fluorescent dye from nanoparticles via spontaneous or enzyme-catalyzed reactions in vivo. The changes in the nanoparticle sizes and shapes were investigated in an HCl solution for 6 hours. RITC-SiNPs were orally administered to healthy nude mice at a dose of 100 mg/kg. Optical imaging studies were performed at 2, 4, and 6 hours after oral administration. The mice were sacrificed at 2, 4, 6, and 10 hours post-administration, and ex vivo imaging studies were

We propose a new neurally-inspired model that can learn to encode the global relationship context of visual events across time and space and to use the contextual information to modulate the analysis by synthesis process in a predictive coding framework. The model learns latent contextual representations by maximizing the predictability of visual events based on local and global contextual information through both top-down and bottom-up processes. In contrast to standard predictive coding models, the prediction error in this model is used to update the contextual representation but does not alter the feedforward input for the next layer, and is thus more consistent with neurophysiological observations. We establish the computational feasibility of this model by demonstrating its ability in several aspects. We show that our model can outperform state-of-art performances of gated Boltzmann machines (GBM) in estimation of contextual information. Our model can also interpolate missing events or predict future events in image sequences while simultaneously estimating contextual information. We show it achieves state-of-art performances in terms of prediction accuracy in a variety of tasks and possesses the ability to interpolate missing frames, a function that is lacking in GBM.

The multidisciplinary goal was to develop an integrated conceptualization of the mid-level encoding of 3D object structure from multiple surface cues. Psychophysical studies showed that depth continuity is a prerequisite for facilitation of Gabor target detection in the context of flanking Gabors, and that, similarly, surface continuity in purely disparity-defined slanted surfaces was strongly enhanced in distributed patch detections as a function of stimulus duration in this dual discrimination task. The unprecedented dips of performance reduction in the component psychometric functions was captured in a computational model based on a novel Leaky Drift Diffusion Theory that we developed for the underlying neural signals, which can serve as an analytic basis for the time course of all neural decision processes. The time course of depth surface perception was studied in a coordinated trio of psychophysical, neurophysiological and functional imaging studies, showing that the perceptual processing of disparity and integration of 3D surface information across depth cues has time courses of several seconds, attesting to complexity of the neural processing hardware. Three complementary computational modeling projects from three collaborating laboratories showed how surface reconstruction could be accomplished across the typically sparse depth information available, and integrated among sparse, incommensurate cue modalities.

This is the report of a study to further investigate the fabrication of thin films of the compound semiconductor Ag-In-Zn-S. The researchers also measured the photoelectrochemical activities of these compounds. By using electrophoretic deposition, the annealed powders of this quaternary solid solution were deposited on the ITO substrate. A larger photo-current density was measured by illuminating on the glass side when the film is thick. At 0 V v.s. SCE, the photo-current density is 0.137 mA/square cm. SEM images show that the structure of the film is porous. The exposed area on ITO is expected to be the reactive sites when bias voltage is applied, with or without light irradiation. In a separate experiment, the metal oxide (SiO2 or SnO2) -coated nanoshells were synthesized. SnO2-coated nanoshells provided enhanced stability at high pH. The effect of temperature and localized surface plasmon resonance, typically with addition of Ag@Au NPs (nanoparticles) was evaluated. Layered structure composed of nanoshell/SiO2/ZnO was also studied. With/without silica layer, nanoshells (NSs) show totally different absorption peaks appeared in UV-vis spectrum. With the presence of nanoshells@SiO2, the IR range in the solar spectrum can be fully utilized. The incorporation of nanoshell@SiO2, as well as the nanostructure of photocatalyst will be further investigated.

L1 ligase (L1L) molecular switch is an in vitro optimized synthetic allosteric ribozyme that catalyzes the regioselective formation of a 5'-to-3' phosphodiester bond, a reaction for which there is no known naturally occurring RNA catalyst. L1L serves as a proof of principle that RNA can catalyze a critical reaction for prebiotic RNA self-replication according to the RNA world hypothesis. L1L crystal structure captures two distinct conformations that differ by a reorientation of one of the stems by around 80Å and are presumed to correspond to the active and inactive state, respectively. It is of great interest to understand the nature of these two states in solution and the pathway for their interconversion. In this study, we use explicit solvent molecular simulation together with a novel enhanced sampling method that utilizes concepts from network theory to map out the conformational transition between active and inactive states of L1L. We find that the overall switching mechanism can be described as a three-state/two-step process. The first step involves a large-amplitude swing that reorients stem C. The second step involves the allosteric activation of the catalytic site through distant contacts with stem C. Using a conformational space network representation of the L1L switch transition, it is shown that the connection between the three states follows different topographical patterns: the stem C swing step passes through a narrow region of the conformational space network, whereas the allosteric activation step covers a much wider region and a more diverse set of pathways through the network.

Horizontal connections in the primary visual cortex have been hypothesized to play a number of computational roles: association field for contour completion, surface interpolation, surround suppression, and saliency computation. Here, we argue that horizontal connections might also serve a critical role for computing the appropriate codes for image representation. That the early visual cortex or V1 explicitly represents the image we perceive has been a common assumption in computational theories of efficient coding (Olshausen and Field (1996)), yet such a framework for understanding the circuitry in V1 has not been seriously entertained in the neurophysiological community. In fact, a number of recent fMRI and neurophysiological studies cast doubt on the neural validity of such an isomorphic representation (Cornelissen et al., 2006; von der Heydt et al., 2003). In this study, we investigated, neurophysiologically, how V1 neurons respond to uniform color surfaces and show that spiking activities of neurons can be decomposed into three components: a bottom-up feedforward input, an articulation of color tuning and a contextual modulation signal that is inversely proportional to the distance away from the bounding contrast border. We demonstrate through computational simulations that the behaviors of a model for image representation are consistent with many aspects of our neural observations. We conclude that the hypothesis of isomorphic representation of images in V1 remains viable and this hypothesis suggests an additional new interpretation of the functional roles of horizontal connections in the primary visual cortex.

Probabilistic generative modeling of data distributions can potentially exploit hidden information which is useful for discriminative classification. This observation has motivated the development of approaches that couple generative and discriminative models for classification. In this paper, we propose a new approach to couple generative and discriminative models in an unified framework based on PAC-Bayes risk theory. We first derive the model-parameter-independent stochastic feature mapping from a practical MAP classifier operating on generative models. Then we construct a linear stochastic classifier equipped with the feature mapping, and derive the explicit PAC-Bayes risk bounds for such classifier for both supervised and semi-supervised learning. Minimizing the risk bound, using an EM-like iterative procedure, results in a new posterior over hidden variables (E-step) and the update rules of model parameters (M-step). The derivation of the posterior is always feasible due to the way of equipping feature mapping and the explicit form of bounding risk. The derived posterior allows the tuning of generative models and subsequently the feature mappings for better classification. The derived update rules of the model parameters are same to those of the uncoupled models as the feature mapping is model-parameter-independent. Our experiments show that the coupling between data modeling generative model and the discriminative classifier via a stochastic feature mapping in this framework leads to a general classification tool with state-of-the-art performance., Comment: 6 pages, 3 figures

Preparation of next-generation photocatalysts is presented. The indium content in this four-component photocatalyst was found to play a critical role. The highest hydrogen evolution rate is 970 micrometersol/h.g. The [Zn]/[Ag] also determined the utilization of visible-light region in the solar spectrum. Samples with various [Zn]/[Ag] compositions show different photocatalytic activities in full spectrum and in visible light. Secondly, it was demonstrated that at an elevated temperature, the hydrogen evolution rate is higher. Future study will focus on improving the efficiency of our second-generation photocatalyst, create various core-shell structures, measure the hydrogen production rate, and test the stability of this composite system.

This study was an attempt to prove that the thickness of the outer egg membrane is the key factor that contributes to the normal zoeal larvae hatching rate of the in vitro artificially fertilized eggs of the Japanese mitten crab Eriocheir japonica by artificially stretching and thinning the outer egg membrane with the help of water surface tension. The results showed that the artificially thinned outer egg membrane became six times thinner. In addition, the normal zoeal larvae hatching rate of the in vitro artificially fertilized eggs rose from 10% to over 66% after the thinning treatment. Discussion focuses on the mechanism of this thinning phenomenon, the role the thinned outer egg membrane plays in the increase of the normal zoeal larvae hatching rate of the in vitro artificially fertilized eggs and the utilization of this particular outer egg membrane thinning method.

The L1 ligase is an in vitro selected ribozyme that uses a noncanonically base-paired ligation site to catalyze regioselectively and regiospecifically the 5' to 3' phosphodiester bond ligation, a reaction relevant to origin of life hypotheses that invoke an RNA world scenario. The L1 ligase crystal structure revealed two different conformational states that were proposed to represent the active and inactive forms. It remains an open question as to what degree these two conformers persist as stable conformational intermediates in solution, and along what pathway are they able to interconvert. To explore these questions, we have performed a series of molecular dynamics simulations in explicit solvent of the inactive-active conformational switch in L1 ligase. Four simulations were performed departing from both conformers in both the reactant and product states, in addition to a simulation where local unfolding in the active state was induced. From these simulations, along with crystallographic data, a set of four virtual torsion angles that span two evolutionarily conserved and restricted regions were identified as dynamical hinge points in the conformational switch transition. The ligation site visits three distinct states characterized by hydrogen bond patterns that are correlated with the formation of specific contacts that may promote catalysis. The insights gained from these simulations contribute to a more detailed understanding of the coupled catalytic/conformational switch mechanism of L1 ligase that may facilitate the design and engineering of new catalytic riboswitches.

This study shows that by removing the outer membranes of the in vitro artificially fertilized eggs of the Japanese mitten crab Eriocheir japonica, the normal zoea hatching rate of the eggs can be raised from 10% up to over 90%. This finding suggests that the normal zoea hatching rate is strongly influenced by the outer egg membranes of the in vitro artificially fertilized eggs. In addition, observations by transmission electron microscopy indicate that the outer egg membranes of the in vitro artificially fertilized eggs are 1.5 to 3 times thicker than those of the naturally spawned eggs. Discussion focuses on the role these thick outer egg membranes play in the normal zoea hatching rate of in vitro artificially fertilized eggs.

This paper describes an electronic system for obtaining mature, unfertilized eggs from decapods. The system was a by-product of the author’s many years of research on in vitro artificial fertilization in crabs (brachyurans). Based on the results of multiple tests, the system has proved to be highly utilizable for research involving the use of mature, unfertilized eggs. The greatest advantage of this system is that it saves researchers a tremendous amount of time and trouble, as the collection of mature, unfertilized eggs is an extremely exhausting procedure that requires 24-hour monitoring over a period of days or weeks. Although the system was originally designed for research with the fresh water mitten crab Eriocheir japonica, it can be used with other decapods with slight modifications.

This paper describes an electronic system for obtaining mature, unfertilized eggs from decapods. The system was a by-product of the author’s many years of research on in vitro artificial fertilization in crabs (brachyurans). Based on the results of multiple tests, the system has proved to be highly utilizable for research involving the use of mature, unfertilized eggs. The greatest advantage of this system is that it saves researchers a tremendous amount of time and trouble, as the collection of mature, unfertilized eggs is an extremely exhausting procedure that requires 24-hour monitoring over a period of days or weeks. Although the system was originally designed for research with the fresh water mitten crab Eriocheir japonica, it can be used with other decapods with slight modifications.

Although the hammerhead ribozyme is regarded as a prototype for understanding RNA catalysis, the mechanistic roles of associated metal ions and water molecules in the cleavage reaction remain controversial. We have investigated the catalytic potential of observed divalent metal ions and water molecules bound to a 2 A structure of the full-length hammerhead ribozyme by using X-ray crystallography in combination with molecular dynamics simulations. A single Mn(2+) is observed to bind directly to the A9 phosphate in the active site, accompanying a hydrogen-bond network involving a well-ordered water molecule spanning N1 of G12 (the general base) and 2'-O of G8 (previously implicated in general acid catalysis) that we propose, based on molecular dynamics calculations, facilitates proton transfer in the cleavage reaction. Phosphate-bridging metal interactions and other mechanistic hypotheses are also tested with this approach.

The chromosomes of Crustacea, especially decapods, are usually numerous, small in size, and punctiform. These characteristics make it difficult for researchers to further investigate the chromosomes by light microscopy. As an alternative, the scanning electron microscopy (SEM) is recommended because it has higher resolution and is able to provide valuable three-dimensional data and X-ray microanalysis. The aim of this study is to obtain some fundamental knowledge for the preparation of the chromosomes of crustaceans for SEM observations. Since crustacean chromosomes are so small, impurities such as mucus and cytoplasm easily obscure the chromosome surface making it impossible for researchers to observe their fine structure. We used Artemia parthenogenetica as a model and confirmed impurities originating from the nauplii paste and the cytoplasm of the cells, and demonstrate that the nauplii paste can be removed by rinsing and the cytoplasm can be digested by trypsin.

Thesis (M.S.)--Massachusetts Institute of Technology, Dept. of Chemical Engineering, 1995., Includes bibliographical references (leaves 94-98)., by Tai-Gyu Lee., M.S.

This study descripts processing of biphasic calcium-phosphate (BCP) and poly-L-lactide (PLLA) biocomposite implant material. The composite was obtained by mixing completely dissolved PLLA with granules of high crystalline BCP and was compacted by hot pressing using cylindrical dies at 450 K temperature and 98.1 MPa pressure, for 30 and 60 minutes. Wide-angle Xray structural (WAXS) analyses of BCP, PLLA and BCP/PLLA composite blocks were made followed by calorimetric (DSC) tests in the 320-520 K temperature range. Compression tests revealed that Young’s modulus and compressive strength of the composite increased with extended hot pressing time and were found to be within the bounds of the cortical bone values.

Research and development of the Weather Systems Processor (WSP) is being accomplished in an effort to satisfy one of the Federal Aviation Administration's (FAA) performance goals in the area of System Efficiency, which is to develop and demonstrate the capability of new systems to decrease the rate of delays due to weather by 10 percent by the year 2002. This plan addresses the Massachusetts Institute of Technology/Lincoln Laboratory (MIT/LL) Demonstration Validation (DEMVAL) of the test bed WSP at their Terminal Radar Development Facility (TRDF) in Albuquerque, New Mexico. This DEMVAL began in November 1995, and will continue through 1997. The purpose of this plan is to provide an overview of the MIT/LL DEMVAL to assist the FAA with its DEMVAL monitoring responsibilities, and with future WSP Test and Evaluation (T&E) activities, including Development Test and Evaluation (DT&E) and Production Acceptance Test and Evaluation (PAT&E) monitoring and System Test conduct.

This Test and Evaluation Master Plan (TEMP) describes technical and operational testing requirements, general methodology, and responsibilities for the comprehensive system testing of the Weather System Processor (WSP) National Airspace System (NAS) subsystem. This TEMP further establishes an agreement between the developing organization, the user, and the tester to support acquisition decisions, by identifying areas of technical and operational risk, by defining a comprehensive plan to address and resolve the risk, and by providing a structure for reporting the results. The WSP Test and Evaluation Program will ensure that the WSP fulfills the Mission Needs Statement (MNS), meets the requirements in the WSP Requirements Document (RD), applicable NAS requirements, the WSP Specification, applicable Interface Requirements Documents (IRDs), and relevant contractor generated documents that have been approved by the Government.

This document is the Final Report for Operational Test and Evaluation (OT&E) Integration and OT&E Operational testing of the Terminal Doppler Weather Radar (TDWR). It provides a complete account of the results of OT&E testing as well as recommendations for future considerations. This report contains a description of testing and evaluation activities including location, participants, specialized test equipment used during testing, test objectives/criteria, test descriptions, test results, and methods used for data collection and analysis. This report also contains a comprehensive discussion of overall test results. Finally, the report ends with test conclusions and recommendations to improve system performance.

Numerical algorithms are developed to estimate petroleum reservoir properties such as absolute permeability, porosity, and relative permeabilities based on the noisy pressure and flow data. Regularization and spline approximation of spatially varying parameters are employed to convert the ill-posed nature of the problem to a well-posed one. A stabilizing functional with gradient operator is used to measure the non-smoothness of the parameter estimates. The number of spline coefficients along each spatial direction is chosen to be as much as the number of meshes for the reservoir PDE's. New history matching algorithms are developed that determine the regularization parameter during the computation without requiring a priori information and improve the parameter estimates stepwise. A partial conjugate gradient method is employed for the estimation of a single set of parameters, and the steepest descent algorithm is used for the simultaneous estimation of absolute and relative permeabilities. A rough parametric sensitivity analysis is carried out for the simultaneous estimation to improve the convergence. Numerical tests are carried out to estimate the parameters in single- and two-phase reservoirs for the different choices of the stabilizing functionals, the regularization parameters, and the degrees of spline approximation; and the effects of the observation error, the observation time, and the configuration of the observation points are investigated. The results show that the new algorithms generate better parameter estimates over the various possible choices of the estimation conditions.

A wind shear detection system, developed by the Wave Propagation Laboratory (WPL) to operate with the Federal Aviation Administration (FAA) Airport Surveillance Radar ASR-(8), was installed and tested at the FAA Technical Center. Initial tests consisted of hardware and software shakedown and feasibility determinations. Second phase tests compared radar with aircraft and tower winds, evaluated the wind shear measurement capability under various weather conditions, and investigated the effectiveness of a simple two-azimuth pointing strategy. Final efforts consisted of observations in all-weather regimes and tests of a modified velocity-azimuth display (VAD) and a glide slope scan. Results showed the system to be compatible with and to operate satisfactorily with the ASR-8. The processing and spectral display of clear air and precipitation returns is feasible. The accuracy of agreement between radar-measured winds and components of the aircraft-measured winds in both radially oriented flights and runway offset flights using a two-azimuth pointing technique, a glide slope scan, and a modified VAD was examined. Radar versus tower wind agreement was also examined. Potentially dangerous wind shears associated with weather during these tests were detectable. Certain system limitations were also defined and considered. (Author)

The estimation of absolute and relative permeabilities for petroleum reservoirs on the basis of noisy data at wells is considered. The spatially varying absolute permeability is estimated by regularisation combined with a bi-cubic spline approximation. Relative permeability is represented by a given function of saturation with unknown coefficients. Numerical results provide an indication of the estimability of the two permeabilities in conventional petroleum production operations.

The working face around the periphery of an excavation has been divided into 5-6 sections and a new blastability index proposed to classify each in terms of its potential for overbreaking, on the basis of rock mass description, uniaxial compressive strength, joint plane spacing, joint plane orientation, joint plane aperture and filling material. This classification is then used to determine the distance between contour holes and the charging density, so as to minimise overbreaking. For controlling charge density and improving the function of the explosives, the new deck-charge method has been developed to utilise the air-deck charge effect and detonation transmission in the hole. The method was applied at the KyungChoon tunnel site in South Korea., The working face around the periphery of an excavation has been divided into 5-6 sections and a new blastability index proposed to classify each in terms of its potential for overbreaking, on the basis of rock mass description, uniaxial compressive strength, joint plane spacing, joint plane orientation, joint plane aperture and filling material. This classification is then used to determine the distance between contour holes and the charging density, so as to minimise overbreaking. For controlling charge density and improving the function of the explosives, the new deck-charge method has been developed to utilise the air-deck charge effect and detonation transmission in the hole. The method was applied at the KyungChoon tunnel site in South Korea.