CUNY Graduate Center

Author Archives: Allan

CSSW 2016 Call For Posters

March 25th, 2016 | Posted by Allan in Uncategorized - (1 Comments)

The Computer Science Students’ Association is officially issuing a call for posters for the Spring 2016 Computer Science Student Workshop.

The deadline for poster title and abstract submissions is April 18th.
The workshop will be on Thursday, May 5th 2:00pm – 4:00pm.

Register and add food prefences here:

The workshop is an opportunity for students to discuss their research across disciplines and prepare upcoming conference presentations. Presentation of past work or course projects is also encouraged.

You can find information on previous years’ workshops at

Please consider participating! Food and beverages will be provided!

Poster Presenters 2015

May 11th, 2015 | Posted by Allan in Uncategorized - (0 Comments)

Poster Presenters for the Graduate Center Computer Science Student Workshop 2015

Poster Abstracts

Exploring Semantic Representations for a Textual Entailment System
Michelle Renee Morales

Given two sentences (T and H) a Recognizing Textual Entailment (RTE) computational system must determine whether T entails H. Certain linguistic phenomena complicate this task. One property researchers have considered is the effects of monotonicity. Some work has investigated whether monotonicity properties can be learned automatically. In addition, there exists a growing body of research supporting the ability of word vectors to capture semantic properties of words. This work builds upon these ideas and introduces a novel approach to learning downward entailing operators using word vector representations. We propose an unsupervised method for automatically detecting downward entailing operators (DEOs). We compare our approach with two previous approaches: Cheung and Penn (2012) and Danescu et al. (2009). Similar to previous work, we leverage the linguistic knowledge that negative polarity items (NPIs) tend to appear in DE contexts to help automatically detect DEOs from text. Different than previous approaches, our method is the first to use word vectors and a similarity measure to rank possible DEOs. Moreover, we are the first to explore the effects the size of a corpus may have on detection. We find our approach outperforms previous approaches in precision and is the least affected by data size.

Reservoir Computing for Recurrent Neural Networks
Felix Grezes

Historically, research has focused on Recurrent Neural Networks (RNNs) because of their capacity to model dynamical system and/or their biological plausibility. However traditional neural network approaches, such has gradient descent, are either impossible or too slow to be applied successfully to RNNs. The Reservoir Computing paradigm, by separating the optimization of the network layer from the output layer, attempts to solve these issues. Since 2001, it has proven itself to be fast and effective in a variety of applications.

Will My Paper Be Influential? Determining and Predicting Citation Patterns
David Guy Brizan, Kevin Gallagher, Arnab Jahanghir, Theodore Brown

Based on citations, we define the influence of publications in three ways: the count of citations, the diversity of research areas of the citations and the longevity of citations. We then rank publications within a cohort of year and journal in order to normalize and define the most influential by rank. Using this technique, we take a machine learning approach — extract a number of features to predict the most influential — and we find that we are consistently able to predict the most influential publications above baseline (chance) performance.

Classification of Vehicle Parts in Unstructured 3D Point Clouds
Allan Zelener, Phillipos Mordohai, Ioannis Stamos

Unprecedented amounts of 3D data can be acquired in urban environments, but their use for scene understanding is challenging due to varying data resolution and variability of objects in the same class. An additional challenge is due to the nature of the point clouds themselves, since they lack detailed geometric or semantic information that would aid scene understanding. In this paper we present a general algorithm for segmenting and jointly classifying object parts and the object itself. Our pipeline consists of local feature extraction, robust RANSAC part segmentation, part-level feature extraction, a structured model for parts in objects, and classification using state-of-the-art classifiers. We have tested this pipeline in a very challenging dataset that consists of real world scans of vehicles. Our contributions include the development of a segmentation and classification pipeline for objects and their parts; and a method for segmentation that is robust to the complexity of unstructured 3D point clouds, as well as a part ordering strategy for the sequential structured model and a joint feature representation between object parts.

Finish Your Haribo Before You Speak: Detecting Eating Events To Classify Food Type From Speech
Min Ma, Guozhen An, Rachel Rakov, Michael J Madden, Andrew Rosenberg

Although research has been done in classifying common conditions under which speech is produced including speaker-state detection such as intoxication detection and automatic detection of laughter/filled pauses, there has been almost no research differentiating speech produced under multiple eating conditions. In this paper, we take a two-stage framework to deal with a 7-class (6 food types plus “No-Food”) classification problem. We first extend the baseline features to distinguish audio containing speech-while-eating from No-Food. We then sub-categorize speech-while-eating by first detecting eat-ing events, distinguishing “Eating-Only” and “Speaking-While- Eating” segments. We then classify each utterance using a combination of utterance-level acoustic features and segment-level food type predictions, exploring multiple combination strategies.

Automatic Recognition of Unified Parkinson’s Disease Rating from Speech with Acoustic, i-Vector and Phonotactic Features
Guozhen An, David Guy Brizan, Hernisa Kacorri, Min Ma, Michelle Morales, Ali Raza Syed, Andrew Rosenberg

Parkinson’s Disease is a neurodegenerative disease affecting millions of people globally, most of whom present difficulties producing speech sounds. In this paper, we describe a system to identify the degree to which a person suffers from the disease. We use a number of automatic phone recognition-based features and we augment these with i-vector features and utterance- level acoustic aggregations. On the Interspeech 2015 ComParE challenge corpus, we find that these features allow for prediction well above the challenge baseline, particularly under cross- validation evaluation.

Photos from CSSW 2014

March 2nd, 2015 | Posted by Allan in Uncategorized - (0 Comments)

Some photos taken from our student workshop from last year are now available here:

Poster Presenters 2014

March 17th, 2014 | Posted by Allan in Uncategorized - (0 Comments)

Poster Presenters for the Graduate Center Computer Science Student Workshop 2014

Poster Abstracts

Automatic Kronecker Product Model Based Detection of Repeated Patterns in 2D Urban Images by Juan Liu
Repeated patterns (such as windows, tiles, balconies and doors) are prominent and significant features in urban scenes. Therefore, detection of these repeated patterns becomes very important for city scene analysis. This paper attacks the problem of repeated patterns detection in a precise, efficient and automatic way, by combining traditional feature extraction followed by a Kronecker product low- rank modeling approach. Our method is tailored for 2D images of building facades. We have developed algorithms for automatic selection of a representative texture within facade images using vanishing points and Harris corners. After rectifying the input images, we describe novel algorithms that extract repeated patterns by using Kronecker product based modeling that is based on a solid theoretical foundation. Our approach is unique and has not ever been used for facade analysis. We have tested our algorithms in a large set of images.

Knowledge Discovery and Predictive Modeling of Protein-drug Binding Kinetics by Integrating Machine Learning and Molecular Dynamics Simulation by See Hong Chiu
Machine learning provides a cost-effective and timesaving approach to discovering hidden patterns and predicting significant outcomes from data. It has succeeded in predicting binding affinity of novel ligand-protein interactions in equilibrium conditions. However, few works have been done to study ligand-protein unbinding events, which are directly correlated with the biological activity of ligands. In this paper, we combine mechanism-based physical modeling and machine learning to predict drug-target residence time, which is a measure how long a drug will bind to its receptor in non-equilibrium conditions. Using HIV protease inhibitors as an example, we identified and selected features based on decomposed residue pair-wise energy and normal model changes upon ligand bindings.

Surveillance Event Detection System by Chucai Yi
We present a general event detection system evaluated by the Surveillance Event Detection (SED) task of TRECVID 2012 campaign. The proposed system is evaluated on all the seven event categories of the SED task. In our system, a sliding temporal window is employed as the detection unit, which is represented by a histogram of spatial-temporal features including STIP-HOG/HOF and SURF/MHI-HOG. We also investigate the spatial priors of various events by estimating spatial distributions of actions under different camera views in the training data. As non-linear SVMs usually have superior performances but in general are much slower in both training and testing, we therefore employ explicit feature maps to approximate large scale non-linear SVMs by linear ones. In order to deal with highly imbalanced data, our system performs detection by a set of cascade linear SVMs that are learned corresponding to specific events and camera views.

Where am I? -Indoor Localization and Navigation for the Visually-impaired People by Feng Hu
In this poster we propose a localization and navigation system for visually-impaired people in indoor environment with portable omnidirectional lens mounted on smart phone a nd remote GPU-enabled server. Concise short omnidirectional video features are extracted and represented in the smart phone front end, and transmitted to the server end, where database is built and great computation capability is provided. Vertical lines features from HSI  and HSI gradient space are used as video clips representation and key-words searched in database for coarse localization. Refined navigation information are obtained from camera pose estimation and moving parameter estimation process. Data parallelism and task parallelism property are observed from the database and estimation process, thus task can be accelerated by GPU. Experiments on synthetic data and real data are carried out and demonstrate the fastness and robustness of the system.

Sensor Selection Problem by Nooreddin Naghibolhosseini
The sensor network is a group of special objects that can sense the environment and are able to produce spacial information about the environment. Each sensor in a sensor network is an independent device with some energy resource and some information gathering mechanism and transmitting capability. The information gathering mechanism can be any mechanism that can sense the environment and provide information. For example a sensor might be a camera that can see certain range and can cover center areas. The sensor might be a wireless device enable of sensing a some object in some distance or maybe something that can measure the temperature or humidity of the environment. A sensor might be a person with access to a telephone line. A sensor network can be made by putting multiple sensors on an environment and these sensors can collaborate with each other to perform better in the sensor network.
In our model of the sensor network. Each sensor has a cost and a value. and We have a budget to select number of sensors in the environment. The interesting subject in the sensor network is the capability of prediction. Each sensor in the sensor network may predict some other sensors with some probability. For example if we use humidity sensor network then sensor A may predict the value of the sensor B in 20 percent of times. This means if the sensor A shows high humidity sensor B may show the high humidity in 20% of times. The value of sensor A and sensor B might be different because maybe the information of location of the sensor A is more important than the information of the location of the sensor B. We have evaluate some algorithms and found new results.

Sustenance against RL-based Sybil attacks in Cognitive Radio Networks using Dynamic Reputation System by Kenneth Ezirim
Sybil attacks are known form of denial-of-service attacks that are common-place in Dynamic Spectrum Access networks. In this paper, we formulate novel threat and defense mechanisms for the Sybil attack problem in Cognitive Radio Net- works (CRN). We present potential identity sampling strategies that a malicious Sybil attacker can use to enhance its attack capability and impact without being detected. We investigate how a Sybil attacker can leverage reinforced learning to improve its performance. We also formulate a novel dynamic reputation mechanism to defend against such threat that relies on the nodes’ reporting in an intelligent and adaptive manner. Results obtained shows that a Sybil attacker can improve its performance using RL learning technique. It also demonstrates that the use of the dynamic reputation mechanism can considerably reduces the effectiveness of Sybil attacks and improve the accuracy of spectrum decisions.

Using a Topological Descriptor to Investigate Structures of Virus Particles by Lucas Oliveira
An understanding of the three-dimensional structure of a biological macromolecular complex is essential to fully understand its function. A component tree is a topological and geometric image descriptor that captures information regarding the structure of an image based on the connected components determined by different grayness thresholds. We believe interactive visual exploration of component trees of (the density maps of) macromolecular complexes
can yield much information about their structure. To illustrate how component trees can convey important structural information, we consider component trees of four recombinant mutants of the procapsid of a bacteriophage (cystovirus phi6),
and show how differences between the component trees reflect the fact that each non-wild-type mutant of the procapsid has an incomplete set of constituent proteins.

Computer Science Student Workshop 2014

January 21st, 2014 | Posted by Allan in conferences | cssw | events - (2 Comments)

Register now: Click Here.

We are pleased to announce the second annual Computer Science Student Workshop will be on March 21st 11:30 – 4:30 at the Graduate Center in room 5414.

Snacks, lunch, and refreshments will be served!

Come and present your latest work, share ideas, and discuss research with other Computer Science students at the Graduate Center! The event is an excellent opportunity to practice communicating your research in a low-pressure environment. When the time comes for your second exam, dissertation, or conference presentation you’ll feel more prepared. Even if you’re not presenting, still come to support your peers and see what’s happening in other fields.

Register now: Click Here.
Poster submission deadline: March 10th


11:30am – Noon

Opening Ceremonies

Noon – 1:30pm

Poster Session 1

1:00pm – 2:30pm

(Free) Lunch & Student meeting

2:30pm – 4:00pm

Poster Session 2

4:00pm – 4:30pm

Closing Ceremonies

 OpenCUNY » login | join | terms | activity 

 Supported by the CUNY Doctoral Students Council.