Saurabh Singh

Saurabh Singh
Are you Saurabh Singh?

Claim your profile, edit publications, add additional information:

Contact Details

Saurabh Singh

Pubs By Year

Pub Categories

Computer Science - Computer Vision and Pattern Recognition (8)
Physics - Strongly Correlated Electrons (4)
Computer Science - Learning (3)
Mathematics - Number Theory (2)
Physics - Materials Science (2)
Cosmology and Nongalactic Astrophysics (2)
Instrumentation and Methods for Astrophysics (2)
Computer Science - Artificial Intelligence (2)
Computer Science - Neural and Evolutionary Computing (2)
Earth and Planetary Astrophysics (1)
Physics - Superconductivity (1)
Physics - Instrumentation and Detectors (1)
Computer Science - Robotics (1)
Computer Science - Computation and Language (1)
Statistics - Machine Learning (1)
Physics - General Physics (1)

Publications Authored By Saurabh Singh

We introduce a stop-code tolerant (SCT) approach to training recurrent convolutional neural networks for lossy image compression. Our methods introduce a multi-pass training method to combine the training goals of high-quality reconstructions in areas around stop-code masking as well as in highly-detailed areas. These methods lead to lower true bitrates for a given recursion count, both pre- and post-entropy coding, even using unstructured LZ77 code compression. Read More

A simple and low cost apparatus has been designed and built to measure the electrical resistivity, ($\rho$), of metal and semiconductors in 300-620 K temperature range. The present design is suitable to do measurement on rectangular bar sample by using conventional four-probe dc method. A small heater is made on the sample mounting copper block to achieve the desired temperature. Read More

A grand goal of computer vision is to build systems that learn visual representations over time that can be applied to many tasks. In this paper, we investigate a vision-language embedding as a core representation and show that it leads to better cross-task transfer than standard multi-task learning. In particular, the task of visual recognition is aligned to the task of visual question answering by forcing each to use the same word-region embeddings. Read More

We propose a method for lossy image compression based on recurrent, convolutional neural networks that outperforms BPG (4:2:0 ), WebP, JPEG2000, and JPEG as measured by MS-SSIM. We introduce three improvements over previous research that lead to this state-of-the-art result. First, we show that training with a pixel-wise loss weighted by SSIM increases reconstruction quality according to several metrics. Read More

We address the problem of distance metric learning (DML), defined as learning a distance consistent with a notion of semantic similarity. Traditionally, for this problem supervision is expressed in the form of sets of points that follow an ordinal relationship -- an anchor point $x$ is similar to a set of positive points $Y$, and dissimilar to a set of negative points $Z$, and a loss defined over these distances is minimized. While the specifics of the optimization differ, in this work we collectively call this type of supervision Triplets and all methods that follow this pattern Triplet-Based methods. Read More

Long wavelength spectral distortions in the Cosmic Microwave Background arising from the 21-cm transition in neutral Hydrogen are a key probe of Cosmic Dawn and the Epoch of Reionization. These features may reveal the nature of the first stars and ultra-faint galaxies that transformed the spin temperature and ionization state of the primordial gas. SARAS 2 is a spectral radiometer purposely designed for precision measurement of these monopole or all-sky global 21-cm spectral distortions. Read More

Normally, understanding the temperature dependent transport properties of strongly correlated electron systems remains challenging task due to complex electronic structure and its variations (around E$_{F}$) with temperature. Here, we report the applicability of DFT+U in explaining thermopower ($\alpha$) and electrical conductivity ($\sigma$) in high temperature region. We have measured temperature dependent $\alpha$ and $\sigma$ in the 300-600 K range. Read More

Code-mixing or code-switching are the effortless phenomena of natural switching between two or more languages in a single conversation. Use of a foreign word in a language; however, does not necessarily mean that the speaker is code-switching because often languages borrow lexical items from other languages. If a word is borrowed, it becomes a part of the lexicon of a language; whereas, during code-switching, the speaker is aware that the conversation involves foreign words or phrases. Read More

Let $k\geq 1$ be an integer. Let $\delta_k(n)$ denote the maximum divisor of $n$ which is co-prime to $k$. We study the error term of the general $m$-th Riesz mean of the arithmetical function $\delta_k(n)$ for any positive integer $m \ge 1$, namely the error term $E_m(x)$ where \[ \frac{1}{m!}\sum_{n \leq x}\delta_k(n) \left( 1-\frac{n}{x} \right)^m = M_{m, k}(x) + E_{m, k}(x). Read More

Let $\lambda_i (n)$ $i= 1, 2, 3$ denote the normalised Fourier coefficients of holomorphic eigenform or Maass cusp form. In this paper we shall consider the sum: \[ S:= \frac{1}{H}\sum_{h\leq H} V\left( \frac{h}{H}\right)\sum_{n\leq N} \lambda_1 (n) \lambda_2 (n+h) \lambda_3 (n+ 2h)W\left( \frac{n}{N} \right), \] \noindent where $V$ and $W$ are smooth bump functions, supported on $[1, 2]$. We shall prove a nontrivial upper bound, under the assumption that $H\geq N^{1/2+ \epsilon}$. Read More

In the present work, we show the importance of temperature dependent energy band gap, E$_{g}$(T), in understanding the high temperature thermoelectric (TE) properties of material by considering LaCoO$_{3}$ (LCO) and ZnV$_{2}$O$_{4}$ (ZVO) compounds as a case study. For the fix value of band gap, E$_{g}$, deviation in the values of $\alpha$ has been observed above 360 K and 400 K for LCO and ZVO compounds, respectively. These deviation can be overcomed by consideration of temperature dependent band gap. Read More

Effective SLAM using a single monocular camera is highly preferred due to its simplicity. However, when compared to trajectory planning methods using depth-based SLAM, Monocular SLAM in loop does need additional considerations. One main reason being that for the optimization, in the form of Bundle Adjustment (BA), to be robust, the SLAM system needs to scan the area for a reasonable duration. Read More

We present the thermoelectric (TE) properties of LaCoO$_{3}$ compound in the temperature range 300-600 K. The experimental value of Seebeck coefficient ($\alpha$) at 300 K is found to be $\sim$635 $\mu$V/K. The value of $\alpha$ decreases continuously with increase in temperature and reaches to $\sim$46 $\mu$V/K at $\sim$600 K. Read More

We describe Swapout, a new stochastic training method, that outperforms ResNets of identical network structure yielding impressive results on CIFAR-10 and CIFAR-100. Swapout samples from a rich set of architectures including dropout, stochastic depth and residual architectures as special cases. When viewed as a regularization method swapout not only inhibits co-adaptation of units in a layer, similar to dropout, but also across network layers. Read More

In the present work, we report the experimental thermopower ($\alpha$) data for ZnV$_{2}$O$_{4}$ compound in the high temperature range 300-600 K. The value of $\alpha$ is found to be $\sim$184 and $\sim$126 $\mu$V/K at $\sim$300 and $\sim$600 K, respectively. The temperature dependent behavior of $\alpha$ is almost linear in the measured temperature range. Read More

We present a method that learns to answer visual questions by selecting image regions relevant to the text-based query. Our method exhibits significant improvements in answering questions such as "what color," where it is necessary to evaluate a specific location, and "what room," where it selectively identifies informative image regions. Our model is tested on the VQA dataset which is the largest human-annotated visual question answering dataset to our knowledge. Read More

A simple apparatus for the measurement of Seebeck coefficient ($\alpha$) in the temperature range 300-620 K has been fabricated. Our design is appropriate for the characterization of samples with different geometries like disk and rod shaped. The sample holder assembly of the apparatus has been designed in such a way that, single heater used for sample heating purpose is enough to provide a self maintain temperature gradient (1-10 K) across the sample. Read More

We present a simple deep learning framework to simultaneously predict keypoint locations and their respective visibilities and use those to achieve state-of-the-art performance for fine-grained classification. We show that by conditioning the predictions on object proposals with sufficient image support, our method can do well without complicated spatial reasoning. Instead, inference methods with robustness to outliers, yield state-of-the-art for keypoint localization. Read More

Detection of the global redshifted 21-cm signal is an excellent means of deciphering the physical processes during the Dark Ages and subsequent Epoch of Reionization (EoR). However, detection of this faint monopole is challenging due to high precision required in instrumental calibration and modeling of substantially brighter foregrounds and instrumental systematics. In particular, modeling of receiver noise with mK accuracy and its separation remains a formidable task in experiments aiming to detect the global signal using single-element spectral radiometers. Read More

We investigate the anomalies in the Earth - Moon system using ancient eclipse data. We identify nine groups of anomalous eclipses between 400 and 1800 AD recorded in parts of India that should have completely missed the subcontinent as per NASA simulations (Espenak and Meeus, 2011). We show that the typical correction to the lunar location required to reconcile the anomalous eclipses is relatively small and consistent with the fluctuations in the length of day that are observed in recent periods. Read More

Correlation between structural/microstructural and magneto-thermal transport properties of FeAs-based SmFeAsO and SmFeAsO0.85F0.15 has been studied in detail. Read More

The goal of this paper is to discover a set of discriminative patches which can serve as a fully unsupervised mid-level visual representation. The desired patches need to satisfy two requirements: 1) to be representative, they need to occur frequently enough in the visual world; 2) to be discriminative, they need to be different enough from the rest of the visual world. The patches could correspond to parts, objects, "visual phrases", etc. Read More