Hewlett-Packard Labs Research Articles

Transfer learning based handwritten character recognition of tamil script using inception-V3 Model

The paper describes the excellent method to get first-rate accuracy and performance in the discipline of Tamil character recognition in a handwritten mode. However, the subject is still at a nascent stage and grossly lacks adequate accuracy in the Tamil language, even though several studies have been conducted within the discipline of handwritten character recognition. This paper draws the attention to the offline handwritten recognition for the Tamil language using the Inception-v3 based transfer learning method. The proposed work is conducted on the readily available HP Tamil handwritten character offline dataset (Hewlett-Packard Lab: hpl-tamil-iso-char-offline-1.0.). It reveals that with the suitable arrangement of transfer learning approach with Inception-v3, the pre-trained model can achieve the recognition accuracy of 93.1%, overtaking the former deep learning designs. The achieved accuracy is due to the use of a pre-trained version with transfer learning that regularly hastens the method of the training process on a new task. Overall, this results in higher accuracy and a more capable version.

Editorial: Neuromorphic Memristive Computation: Where Memristor-Based Designs Meet Artificial Intelligence Applications

Intelligence Applications 16The definition of a memristor was introduced by Professor L. O. Chua in 1971. By generalizing the 17 definition of a memristor, memristive devices and systems were also proposed. Since the real 18 memristor was implemented at Hewlett-Packard Labs, there has been an increasing interest in 19 memristor and its applications. Memristor is a kind of synaptic element with nanometer size and continuously variable memristance. 68The bridge synaptic circuit constructed by the memristor has a simple structure and precise control. In 69 practice, because of the non-linear characteristics of memristor, it is not easy to control synaptic circuit 70 and errors in weights appear. Therefore, a novel memristor synaptic circuit is proposed by Wang et al., 71 named the dual-mode memristor bridge synaptic neural network. The proposed method can make the 72 weights more linear by controlling the input voltages and make the outputs more linear by using 73 symmetrical positive and negative pulses. As a result, the proposed synaptic circuit is easier to be 74 controlled. The numerical simulations are conducted and verify the feasibility. Furthermore, the 75 simulation experiments are conducted for edge extraction of grayscale birds' images in the airport for 76 bird recognition applied for the bird repelling applications. 77

NbO2-Mott Memristor: A Circuit- Theoretic Investigation

This paper presents a circuit-theoretic analysis of a NbO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> -Mott memristor fabricated at Hewlett-Packard Labs. It investigates mechanisms behind the origin of complexity based on local activity, which characterizes the behavior of this outstanding nanodevice. We propose an accurate, particularly simplified version of a recently introduced physical model suitable for large-scale circuit simulations. Following the concept of local activity, we then conduct a small-signal circuit-theoretic derivation of the impedance and associated small-signal equivalent circuit elements to analyze device stability and frequency response. Finally, our analysis reveals locally active operating regions, as well as regions where the device dynamics are positioned on the edge of chaos. The latter regions are crucial for designing bio-inspired computing systems.

A Novel Modeling Methodology for Memristive Systems Using Homotopy Perturbation Methods

The physical conception of the memristor by the Hewlett-Packard Labs has opened the possibility of combining this new fundamental circuit element with traditional devices in order to carried out novel applications. An important step for incorporating the memristor into the computer-aided design tools is to find an expression that represents the behavior of the device. In this paper, a novel modeling methodology for memristive systems using homotopy perturbation methods is introduced. This methodology is capable of generating a semi-symbolical expression that represents the memristive behavior of the device. The new memristor model is characterized in order to determinate the impact of the variables on the memristive behavior and establishes the range of compliance using the principal fingerprints of the memristor. Besides, the novel memristor model is depicted by two cases of study. Finally, the advantages of the memristor model are presented and discussed.

Two Memristor SPICE Models and Their Applications in Microwave Devices

This paper presents two simple SPICE circuit models of the memristor using two different kinds of integrators. These models expand and simplify the previous methods of solving the memristor's modeling equations presented by Hewlett–Packard Lab. The behaviors of the two memristor models are investigated when they are excited by a sinusoidal voltage source. Both models satisfy the general features of memristive systems such as having a zero-crossing property in the form of an i–v Lissajous figure. In order to explore the unique characteristics and applications of the memristor in microwave devices, first we incorporate the memristor in a microstrip transmission line as a load. We do the analysis using a finite-difference time-domain simulator integrated with a nonlinear SPICE circuit solver. Furthermore, we design a reconfigurable microstrip bandpass filter based on a memristor-loaded resonator, and utilize a memristor as a carrier-wave modulator connecting the microstrip patch antenna to the ground.

Ivan A. Sag

Ivan A. Sag

Imputation of missing values for semi-supervised data using the proximity in random forests

This paper presents a procedure that imputes missing values by using random forests on semi-supervised data. Applying our method to Hewlett-Packard Lab.’s spam data and Edgar Anderson’s iris data, we found that the rate of correct classification is higher than that of other methods: a simple expansion of Liaw’s ‘rfImpute’ for (un)supervised data and the k-nearest neighbour method (kNN). Our method allows missing predictor variables as well as missing response variable. An imputation that uses random forests for semi-supervised cases in the training dataset has never been implemented until now.

Focus on Authors

Focus on Authors

Memristor based-elements for chaotic circuits

The peculiar features of the memristor, a fundamental passive two-terminal element characterized by nonlinear relationship between charge and flux, promise to revolutionize integrated circuit design in the next few decades. Besides its most popular potential applications, ultra-dense non-volatile memories and brain-simulating systems, much research has been lately devoted to their use in chaotic circuits. Although the physical memristor is inherently-asymmetric, complementary behaviors arise in devices with opposite polarity. In this work we demonstrate that this makes it feasible to devise a number of practical realizations of a monotone-increasing odd-symmetric charge-flux nonlinearity suitable for chaos-based applications through the sole use of physical memristors of the kind recently fabricated at Hewlett-Packard Labs. Confirmation for such claim is obtained through comparison of chaotic behavior of two modified Chua's oscillators, in which the nonlinear diode is replaced in one case with an artificial memristor with symmetric nonlinearity and in the other case with one of the proposed symmetric memristor combinations

Q&amp;AHP's running man

Prith Banerjee discusses collaborating with universities, his startup experiences, and Hewlett-Packard Lab's approach to research and development.

Corporate Research and Venture Capital Can Learn from Each Other

OVERVIEW:After carefully evaluating a variety of mechanisms to leverage the world-class venture capital (VC) firms in the vicinity of Hewlett-Packard's Silicon Valley headquarters, Hewlett-Packard Labs and the venture capital firm Foundation Capital entered into a strategic relationship. This relationship focused on the unique needs of a large corporate research organization, and was not intended to replace or substitute for on-going relationships between HP's business organizations and the venture capital community. The relationship has enabled both firms to understand more clearly the fundamental differences in their two different innovation models, and it is suggested that other large central research labs could benefit from a similar relationship. HP Labs, for instance, has become more selective about investments in promising new technologies, and has been able to propose more focused market entry strategies for several of its new technologies. There are also ways to achieve such benefits without actually establishing a formal relationship.

Preface: Volume 22

This volume contains the proceedings of the First International Workshop on Probabilistic Methods in Verification (PROBMIV′98) held 19-20 June 1998 in Indianapolis, Indiana, USA, as a satellite event to the Thirteenth International IEEE Symposium on Logic in Computer Science (LICS′98).Probabilistic reasoning and modelling inherently features in the design processes for software and hardware systems, either as a means to derive efficient solutions (e.g. randomisation), or as a result of unreliable or unpredictable behaviour (e.g. in fault-tolerant systems, computer networks, etc.). Probabilistic verificationencompasses a range of theoretical, algorithmic and programming methods that assist in establishing the correctness of probabilistic systems against specifications. Typically such methods involve calculating the probability bounds for a temporal logic formula being satisfied, based on an appropriate probability space on computations, but also include the analysis of properties with respect to cost and long-run average behaviour. Additionally, probabilistic verification can provide guarantees that specifications hold with satisfactory probability in cases when conventional property verification, such as model checking, is not feasible.This two-day workshop aimed to and succeeded in getting together researchers whose work ranges across the whole spectrum of probabilistic methods, from computational linear algebra and semantics, through probabilistic languages and logics, to model checking algorithms and implementation work. The workshop was held in an informal atmosphere, with lively discussions. Its programme consisted of ten invited lectures, seven accepted submissions and a panel discussion. The invited speakers were:Rajeev Alur(University of Pennsylvania)Luca de Alfaro(University of California at Berkeley)Christel Baier &Vicky Hartonas-Garmhausen(Mannheim University)(Carnegie Mellon University)Jeremy Gunawardena(BRIMS, Hewlett-Packard Labs)Marek Karpinski(University of Bonn)Annabelle McIver(Oxford University)Prakash Panangaden(McGill University)Roberto Segala(University of Bologna)K Narayan Kumar(SUNY Stony Brook)Moshe Vardi(Rice University)The panel discussion focused on the question“What tools and theory are needed in order to make probabilistic verification practical?”It began with a position statement by Moshe Vardi who identified the key reasons for, in his view, the success of conventional model checking in practice, and the challenges that researchers in probabilistic verification have yet to address: efficient data structures, good heuristics, methods for dealing with very large state-space systems, the ability to generate insightful counter-traces if systems fail to meet a specified property, and convincing applications and case studies. Marek Karpinski emphasised one advantage of probabilistic verification: that it can compute epsilon-approximations to otherwise infeasible problems with high probability. Rajeev Alur pointed out the lack of existing tools in this area and encouraged work on probabilistic on-the-fly model checking. Prakash Panangaden argued for the development of a robust mathematical theory for modelling and reasoning about probabilistic systems. He also wanted to see a closer connection of such work with the contributions of performance modelling, which has well developed tools and analysis methodologies. Jeremy Gunawardena encouraged this scientific community to design appropriate languages, models, algorithms and tools that are capable of reasoning about, e.g., internet traffic. It was felt that probabilistic verification will continue to gain in importance because of the many real-world phenomena in need of a probabilistic and stochastic analysis: computer networks, the Ethernet, etc.As we write this foreword, the second PROBMIV workshop, PROBMIV′99, is well underway and there is a Dagstuhl Seminar on this subject planned for May 2000. We thus offer you this collection of papers, and hope you will find it useful and stimulating!The organisers gratefully acknowledge partial support from Basic Research Institute in the Mathematical Sciences (BRIMS), Hewlett-Packard, Bristol, and also the European Commission via Esprit Working Group FIREworks (23531).Christel Baier, Michael Huth (Co-Chair), Marta Kwiatkowska (Co-Chair)Mark Ryan, Guest Editors

Modeling a fibre channel switch with stochastic Petri nets

Gianfranco Ciardo Dept. of Computer Science, College of William and Mary Williamsburg, VA 23187-8795, USA ciardo@cs.wm.edu Ludmila Cherkasova, Vadim Kotov, and Tomas Rokicki Hewlett-Packard Labs 1501 Page Mill Road, Palo Alto, CA 94303, USA fcherkasova,kotov,rokickig@hpl.hp.com As computing systems become increasingly complex and concurrent, performance analysis becomes more important, especially when designing an overall architecture. Simple numerical approximations can often yield bounds on performance. Queueing theory and Markovian analysis can provide insight into the steady-state performance of the system. Discreteevent simulation can provide a more detailed view of the system behavior. We used all of the above techniques in the performance study of a Fibre Channel switch. An approximate stochastic Petri net (SPN) model was employed, obtained by exploiting symmetries in the system. The most challenging aspect was determining transition ring probabilities that accurately re ected the dynamic behavior of the exact SPN. This model was valided against a detailed simulation and found to be accurate within 3%. See [1] for details and references. 1 System description We consider a switch with N ports, numbered from 0 to N 1 (Fig. 1). The switch contains bu er pools organized by input port, a router to transfer header information to output queues, and a scheduler to schedule the transfer of data from an input bu er across the crossbar to an output port. When a message arrives at a port, it is read into a bu er as soon as one is available. The router cyclically polls each port, bypassing ports without new messages. During the poll of a particular port, the router reads the header information of the new message and inserts it into a scheduler table. This table is organized as N FIFO queues, one per output port. The router starts routing a message as soon as the header is in 0 1 N-1 Input Port 0 B slots