Search arXiv⌕ Search

arXiv subjects

Zhuang Liu

Publications and source records attributed to Zhuang Liu.

106 records · Page 6Linked to original sources
arXiv2017

A conjugate gradient method for electronic structure calculations

In this paper, we study a conjugate gradient method for electronic structure calculations. We propose a Hessian based step size strategy, which together with three orthogonality approaches yields three algorithms for computing the ground state energy of atomic and molecular systems. Under some mild assumptions, we prove that our algorithms converge locally. It is shown by our numerical experiments that the conjugate gradient method is efficient.

math.NA↗
arXiv2017

Learning Efficient Convolutional Networks through Network Slimming

The deployment of deep convolutional neural networks (CNNs) in many real world applications is largely hindered by their high computational cost. In this paper, we propose a novel learning scheme for CNNs to simultaneously 1) reduce the model size; 2) decrease the run-time memory footprint; and 3) lower the number of computing operations, without compromising accuracy. This is achieved by enforcing channel-level sparsity in the network in a simple but effective way. Different from many existing approaches, the proposed method directly applies to modern CNN architectures, introduces minimum overhead to the training process, and requires no special software/hardware accelerators for the resulting models. We call our approach network slimming, which takes wide and large networks as input models, but during training insignificant channels are automatically identified and pruned afterwards, yielding thin and compact models with comparable accuracy. We empirically demonstrate the effectiveness of our approach with several state-of-the-art CNN models, including VGGNet, ResNet and DenseNet, on various image classification datasets. For VGGNet, a multi-pass version of network slimming gives a 20x reduction in model size and a 5x reduction in computing operations.

cs.CV↗
arXiv2017

Snapshot Ensembles: Train 1, get M for free

Ensembles of neural networks are known to be much more robust and accurate than individual networks. However, training multiple deep networks for model averaging is computationally expensive. In this paper, we propose a method to obtain the seemingly contradictory goal of ensembling multiple neural networks at no additional training cost. We achieve this goal by training a single neural network, converging to several local minima along its optimization path and saving the model parameters. To obtain repeated rapid convergence, we leverage recent work on cyclic learning rate schedules. The resulting technique, which we refer to as Snapshot Ensembling, is simple, yet surprisingly effective. We show in a series of experiments that our approach is compatible with diverse network architectures and learning tasks. It consistently yields lower error rates than state-of-the-art single models at no additional training cost, and compares favorably with traditional network ensembles. On CIFAR-10 and CIFAR-100 our DenseNet Snapshot Ensembles obtain error rates of 3.4% and 17.4% respectively.

cs.LG↗
arXiv2016

Deep Networks with Stochastic Depth

Very deep convolutional networks with hundreds of layers have led to significant reductions in error on competitive benchmarks. Although the unmatched expressiveness of the many layers can be highly desirable at test time, training very deep networks comes with its own set of challenges. The gradients can vanish, the forward flow often diminishes, and the training time can be painfully slow. To address these problems, we propose stochastic depth, a training procedure that enables the seemingly contradictory setup to train short networks and use deep networks at test time. We start with very deep networks but during training, for each mini-batch, randomly drop a subset of layers and bypass them with the identity function. This simple approach complements the recent success of residual networks. It reduces training time substantially and improves the test error significantly on almost all data sets that we used for evaluation. With stochastic depth we can increase the depth of residual networks even beyond 1200 layers and still yield meaningful improvements in test error (4.91% on CIFAR-10).

cs.LG↗
arXiv2010

Multiplexed five-color molecular imaging of cancer cells and tumor tissues with carbon nanotube Raman tags in the near-infrared

Single-walled carbon nanotubes (SWNTs) with five different C13/C12 isotope compositions and well-separated Raman peaks have been synthesized and conjugated to five targeting ligands in order to impart molecular specificity. Multiplexed Raman imaging of live cells has been carried out by highly specific staining of cells with a five-color mixture of SWNTs. Ex vivo multiplexed Raman imaging of tumor samples uncovers a surprising up-regulation of epidermal growth factor receptor (EGFR) on LS174T colon cancer cells from cell culture to in vivo tumor growth. This is the first time five-color multiplexed molecular imaging has been performed in the near-infrared (NIR) region under a single laser excitation. Near zero interfering background of imaging is achieved due to the sharp Raman peaks unique to nanotubes over the low, smooth autofluorescence background of biological species.

cond-mat.mtrl-sci↗
arXiv2009

Supramolecular Stacking of Doxorubicin on Carbon Nanotubes for in vivo Cancer Therapy

Many therapeutic advantages such as prolonged circulation in the blood, increased tumor drug uptake, enhanced therapeutic efficacy, and markedly reduced toxic side effects are provided by a carbon nanotube based chemotherapeutic formulation. In this system, doxorubicin (DOX) is loaded onto the sidewalls of functionalized single-walled carbon nanotubes by supramolecular pi-pi stacking.

cond-mat.mtrl-sci↗
arXiv2009

Preparation of Carbon Nanotube Bioconjugates for Biomedical Applications

Biomedical applications of carbon nanotubes have attracted much attention in recent years. Here, we summarize our previously developed protocols for functionalization and bioconjugation of single wall carbon nanotubes (SWNTs) for various biomedical applications including biological imaging, sensing and drug delivery. Sonication of SWNTs in solutions of phospholipid-polyethylene glycol (PL-PEG) is our most commonly used protocol of SWNT functionalization. Compared to other frequently used covalent strategies, our non-covalent functionalization protocol largely retains the intrinsic optical properties of SWNTs, which are useful in various biological imaging and sensing applications. Functionalized SWNTs are conjugated with targeting ligands for specific cell labeling in vitro or tumor targeting in vivo. Radio labels are introduced for tracking and imaging of SWNTs in real time in vivo. Moreover, SWNTs can be conjugated with small interfering RNA (siRNA) or loaded with chemotherapy drugs for drug delivery. These procedures take various times ranging from one to five days.

cond-mat.mtrl-sci↗
arXiv2009

PEG Branched Polymer for Functionalization of Nanomaterials with Ultralong Blood Circulation

Nanomaterials have been actively pursued for biological and medical applications in recent years. Here, we report the synthesis of several new poly(ethylene glycol) grafted branched-polymers for functionalization of various nanomaterials including carbon nanotubes, gold nanoparticles (NP) and gold nanorods (NRs), affording high aqueous solubility and stability for these materials. We synthesize different surfactant polymers based upon poly-(g-glutamic acid) (gPGA) and poly(maleic anhydride-alt-1-octadecene) (PMHC18). We use the abundant free carboxylic acid groups of gPGA for attaching lipophilic species such as pyrene or phospholipid, which bind to nanomaterials via robust physisorption. Additionally, the remaining carboxylic acids on gPGA or the amine-reactive anhydrides of PMHC18 are then PEGylated, providing extended hydrophilic groups, affording polymeric amphiphiles. We show that single-walled carbon nanotubes (SWNTs), Au NPs and NRs functionalized by the polymers exhibit high stability in aqueous solutions at different pHs, at elevated temperatures and in serum. Morever, the polymer-coated SWNTs exhibit remarkably long blood circulation (t1/2 22.1 h) upon intravenous injection into mice, far exceeding the previous record of 5.4 h. The ultra-long blood circulation time suggests greatly delayed clearance of nanomaterials by the reticuloendothelial system (RES) of mice, a highly desired property for in vivo applications of nanomaterials, including imaging and drug delivery.

cond-mat.mtrl-sci↗
arXiv2009

Carbon Nanotubes in Biology and Medicine: in vitro and in vivo Detection, Imaging and Drug Delivery

Carbon nanotubes exhibit many unique intrinsic physical and chemical properties and have been intensively explored for biological and biomedical applications. In this review, we summarize the main results of our and other groups in this field and clarify that surface functionalization is critical to the behaviors of carbon nanotubes in biological systems. Ultra-sensitive detection of biological species with carbon nanotubes can be realized after surface passivation to inhibit the non-specific binding of bio-molecules on the hydrophobic nanotube surface. Electrical nanosensors based on nanotubes provide a label-free approach to biological detections. Surface enhanced Raman spectroscopy of CNT opens up a method of protein microarray with down to 1 fM detection sensitivity. In vitro and in vivo toxicity studies reveal that well water soluble and serum stable nanotubes are biocompatible, non-toxic and potentially useful for biomedical applications. In vivo biodistributions vary with the functionalization and possibly also sizes of nanotubes, with a tendency of accumulation in the reticuloendothelial systems, including the liver and spleen, after intravenous administration. If well functionalized, nanotubes may be excreted mainly through the biliary pathway in feces. Carbon nanotube-based drug delivery has shown promises in various in vitro and in vivo experiments including delivery pf small interfering RNA, paclitaxel and doxorubicin. Moreover, SWNTs with various interesting intrinsic optical properties have been used as novel photoluminance, Raman and photoacoustic contrast agents for imaging of cells and animals. Further multidisciplinary explorations in this field are promising and may bring new opportunities to the realm of biomedicine.

cond-mat.mtrl-sci↗
arXiv2008

Protein Microarrays with Carbon Nanotubes as Multi-Color Raman Labels

Detection of biomolecules is important in proteomics and clinical diagnosis and treatment of diseases. Here, we apply functionalized, macromolecular, single walled carbon nanotubes SWNTs as multi-color Raman labels to protein arrays for highly sensitive, multiplexed protein detection. Raman detection utilizes the sharp peaks of SWNTs with minimal background interference, affording a high signal to noise ratio needed for ultra-sensitive detection. Surface-enhanced Raman scattering SERS combined with the strong resonance Raman intensity of SWNTs, affords detection sensitivity down to 1 fM, a three order of magnitude improvement over most of reported fluorescence-based protein detections. We show that human autoantibodies to Proteinase 3 aPR3, a biomarker for the autoimmune disease Wegeners granulomatosis, is detected by Raman in human serum up to a 107 dilution. Moreover, SWNT Raman tags are stable against photobleaching and quenching, and by conjugating different antibodies to pure 12C and 13C SWNT isotopes, we demonstrate two-color SWNT Raman-based protein detection in a multiplexed fashion.

cond-mat.mtrl-sci↗
arXiv2008

Multiplexed Multi-Color Raman Imaging of Live Cells with Isotopically Modified Single Walled Carbon Nanotubes

We show that single walled carbon nanotubes with different isotope compositions exhibit distinct Raman Gband peaks and can be used for multiplexed multi-color Raman imaging of biological systems. Cancer cells with specific receptors are selectively labeled with 3 differently colored SWNTs conjugated with various targeting ligands including Herceptin, anti-Her2, Erbitux, anti-Her1, and RGD peptide, allowing for multi-color Raman imaging of cells in a multiplexed manner. SWNT Raman signals are highly robust against photo-bleaching, allowing long term imaging and tracking. With narrow peak features, SWNT Raman signals are easily differentiated from the auto-fluorescence background. The SWNT Raman excitation and scattering photons are in the near-infrared region, which is the most transparent optical window for biological systems in vitro and in vivo. Thus, SWNTs are novel Raman tags promising for multiplexed biological detection and imaging.

cond-mat.mtrl-sci↗
arXiv2008

Nano-Graphene Oxide for Cellular Imaging and Drug Delivery

Two-dimensional graphene offers interesting electronic, thermal and mechanical properties that are currently explored for advanced electronics, membranes and composites. Here we synthesize and explore the biological application of nano-graphene oxide NGO, single-layer graphene oxide sheets down to a few nanometers in lateral width. We develop functionalization chemistry to impart solubility and compatibility of NGO in biological environments. We obtain size separated pegylated NGO sheets that are soluble in buffers and serum without agglomeration. The NGO sheets are found to be photoluminescent in the visible and infrared regions. The intrinsic photoluminescence of NGO is used for live cell imaging in the near-infrared with little background. We found that simple physisorption via pi-stacking can be used for loading doxorubicin, a widely used cancer drug onto NGO functionalized with antibody for selective cancer cell killing in vitro. Owing to the small size, intrinsic optical properties, large specific surface area,low cost, and useful non-covalent interactions with aromatic drug molecules, NGO is a promising new material for biological and medical applications.

cond-mat.mtrl-sci↗
arXiv2008

Drug delivery with carbon nanotubes for in vivo cancer treatment

Chemically functionalized single-walled carbon nanotubes (SWNTs) have shown promise in tumor targeted accumulation in mice and exhibit biocompatibility, excretion and little toxicity. Here, we demonstrate in-vivo SWNT drug delivery for tumor suppression in mice. We conjugate paclitaxel (PTX), a widely used cancer chemotherapy drug to branched polyethylene-glycol (PEG) chains on SWNTs via a cleavable ester bond to obtain a water soluble SWNT-paclitaxel conjugate (SWNT-PTX). SWNT-PTX affords higher efficacy in suppressing tumor growth than clinical Taxol in a murine 4T1 breast-cancer model, owing to prolonged blood circulation and 10-fold higher tumor PTX uptake by SWNT delivery likely through enhanced permeability and retention (EPR). Drug molecules carried into the reticuloendothelial system are released from SWNTs and excreted via biliary pathway without causing obvious toxic effects to normal organs. Thus, nanotube drug delivery is promising for high treatment efficacy and minimum side effects for future cancer therapy with low drug doses.

cond-mat.mtrl-sci↗
arXiv2008

PEGylated Nano-Graphene Oxide for Delivery of Water Insoluble Cancer Drugs

It is known that many potent, often aromatic drugs are water insoluble, which has hampered their use for disease treatment. In this work, we functionalized nano-graphene oxide (NGO), a novel graphitic material, with branched polyethylene glycol (PEG) to obtain a biocompatible NGO-PEG conjugate stable in various biological solutions, and used them for attaching hydrophobic aromatic molecules including a camptothecin (CPT) analog, SN38 non-covalently via pi-pi stacking. The resulting NGO-PEG-SN38 complex exhibited excellent water solubility while maintaining its high cancer cell killing potency similar to that of the free SN38 molecules in organic solvents. The efficacy of NGO-PEG-SN38 was far higher than that of irinotecan (CPT-11), a FDA approved water soluble SN38 prodrug used for the treatment of colon cancer. Our results showed that graphene is a novel class of material promising for biological applications including future in vivo cancer treatment with various aromatic, low-solubility drugs.

cond-mat.mtrl-sci↗
arXiv2005

Oxidation Resistant Germanium Nanowires: Bulk Synthesis, Long Chain Alkanethiol Functionalization and Langmuir-Blodgett Assembly

A simple method is developed to synthesize gram quantities of uniform Ge nanowires (GeNWs) by chemical vapor deposition on preformed, monodispersed seed-particles loaded onto high surface area silica support. Various chemical functionalization schemes are investigated to passivate the GeNW surfaces using alkanethiols and alkyl Grignard reactions. The stability of functionalization against oxidation of germanium for various alkyl chain lengths is elucidated by X-ray photoelectron spectroscopy. Among all schemes tested, long chain alkanethiols (>=C12) are found to impart the most stable GeNW passivation against oxidation upon extended exposure to ambient air. Further, the chemically functionalized oxidation-resistant nanowires are soluble in organic solvents and can be readily assembled into close-packed Langmuir-Blodgett films potentially useful for future high performance electronic devices.

cond-mat.mtrl-sci↗