More details     In addition to this data source, Visual Studio generates a file that contains the schema for the DataSet class. This file defines the structure of the dataset, including the tables it contains, the columns that are included in each table, the data types of each column, and the constraints that are defined for each table. It is listed in the Solution Explorer window and is given the same name you specified for the dataset in the last step of the Data Source Configuration Wizard with a file extension of xsd. You can view a graphic representation of this schema by double-clicking on this file. Beneath the schema file, the Solution Explorer displays the file that contains the generated code for the DataSet class. In this figure, this code is stored in the MMABooksDataSet.Designer.cs file. When you create bound controls from the data source, the code in this class is used to define the DataSet object that the controls are bound to. Although you may want to View this code to see how it works, you shouldn’t change it. If you do, the dataset may not work correctly. By the way, you should know that a dataset that’s created from a dataset class like the one shown here is called a typed dataset. The code in the dataset class makes it possible for you to refer to the tables, rows, and columns in the typed dataset using the simplified syntax. In contrast, when you use an untyped dataset, you have to refer to the tables, columns, and rows through the collections that contain them. 34
View full slide show




Collection Interface • The Collection interface generalizes the concept of a sequence of elements. • Basic Collection operations include – No-argument and copy constructors – boolean contains( Object x) – returns true if at least one instance of x is in the collection – boolean containsAll( Collection targets) – returns true if all targets are contained in the calling collection object – boolean equals(Object x ) – This is equals for the collection, not the elements in the collection. Intuitive meaning. – Object[ ] toArray( ) – returns an array containing all of the elements – boolean add(T element ) – ensures that the calling collection object contains the specified element. (optional) – boolean addAll( Collection collectionToAdd) – ensures that the calling collection object contains all elements of collectionToAdd (optional) – boolean remove(T element) – removes a single instance of the element from the calling collection object (optional) – boolean removeAll( Collection collectionToRemove) – removes all elements contained in collectionToRemove from the calling collection object (optional) – void clear( ) – removes all elements from the calling collection object (optional)
View full slide show




When to finetune your model? • New dataset is small and similar to original dataset. • train a linear classifier on the CNN codes • New dataset is large and similar to the original dataset • fine-tune through the full network • New dataset is small but very different from the original dataset • SVM classifier from activations somewhere earlier in the network • New dataset is large and very different from the original dataset • fine-tune through the entire network
View full slide show




The List interface extends the Collection interface by supplying some index-oriented methods List-specific behavior that, for example, doesn’t make sense for all Collections class Class Diagram     add(int index, E element) get(int index) set(int index) ... ~ size() : int Some data structures, like queues, are Collections but not Lists – meaning their elements cannot be accessed by index. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Collection «interface» util::List Iterable «interface» util::Collection isEmpty() : boolean contains(Object) : boolean iterator() : Iterator toArray() : Object[] toArray(T[]) : T[] add(E) : boolean remove(Object) : boolean containsAll(Collection) : boolean addAll(Collection) : boolean removeAll(Collection) : boolean retainAll(Collection) : boolean clear() : void equals(Object) : boolean hashCode() : int CS-2851 Dr. Mark L. Hornick ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ size() : int isEmpty() : boolean contains(Object) : boolean iterator() : Iterator toArray() : Object[] toArray(T[]) : T[] add(E) : boolean remove(Object) : boolean containsAll(Collection) : boolean addAll(Collection) : boolean addAll(int, Collection) : boolean removeAll(Collection) : boolean retainAll(Collection) : boolean clear() : void equals(Object) : boolean hashCode() : int get(int) : E set(int, E) : E add(int, E) : void remove(int) : E indexOf(Object) : int lastIndexOf(Object) : int listIterator() : ListIterator listIterator(int) : ListIterator subList(int, int) : List 5
View full slide show




One-handed Keystroke Biometric Identification Competition John V. Monaco1, Gonzalo Perez1, Charles C. Tappert1, Patrick Bours2, Soumik Mondal2, Sudalai Rajkumar3, Aythami Morales4, Julian Fierrez4, Javier Ortega-Garcia4 1. Pace University, 2. Gjøvik University College, 3. Tiger Analytics, 4. Universidad Autónoma de Madrid INTRODUCTION COMPETITION RESULTS WINNING STRATEGIES First place Is it possible for a keystroke biometric system to give accurate results when typing behavior is severely impaired? This competition aimed to answer that question. Participants built classifiers using a labeled keystroke biometric dataset with normal typing behavior only. They then attempted to identify the subjects in an unlabeled dataset that contained some samples that were typed with only one hand. This scenario simulates a severe user handicap. Baseline results indicate a severe degradation in performance for one-handed keystroke samples. Participants had to construct novel classifiers capable of identifying normal and handicapped samples in this competition that ranked the identification accuracy under several different typing conditions. Accuracy for handedness vs. typing condition • Duration features only • Multi-classifier pairwise coupling with 2 regression models and a prediction model • Artificial Neural Network (ANN) • Counter-Propagation Artificial Neural Network (CPANN) • Support Vector Machine (SVM) • Weighted fusion of classifier scores • Features corresponding to the typing condition • Left-side keyboard features for left-hand typing • Right-side keyboard features for right-hand Bottom tree structure for pairwise coupling typing DATA The winning group was awarded a Futronic FS88 Fingerprint Scanner. Three online exams were administered to 64 undergraduate students. Keystrokes were collected using a plugin for Moodle that captures key press and release timestamps on the client and sends this information back to the server. To simulate a typing impairment, students were instructed to • Type normally with both hands on the first exam. • Type with left hand only on the second exam. • Type with right hand only on the third exam. Samples were created by taking 500-keystroke segments separated by at least 50 keystrokes apart. The labeled dataset consisted of one normally-typed sample per student. The unlabeled dataset contained 471 samples from all three typing conditions. Not all of the students in the labeled dataset also appeared in the unlabeled dataset. All samples were provided in millisecond precision and normalized to begin at time 0 to avoid linking the samples by the time the test was taken. Competition participants were allowed to make up to one submission per day, using a plugin for Moodle developed by the authors. Results were automatically scored and remain publicly available: RESEARCH POSTER PRESENTATION DESIGN © 2012 http://biometric-competitions.com/mod/competition/leaderboard.php? www.PosterPresentations.com Accuracy for each typing style Accuracy distribution per student Accuracy vs. typing speed Accuracy distribution per sample for each typing condition Second place • Duration, press-press latency, and release-press latency features • Grouped features based on keyboard layout (left vs. right, top vs. bottom) • Random Forest classifier Third place • Features corresponding to typing condition, similar as above. • Duration, release-press latency, and trigraph features • trigraph features: press to release of alternate keystrokes • Fusion of normalized distance between feature vectors and Least Squares Support Vector Machine (LS-SVM) • Meta-parameters of the LS-SVM determined on an independent dataset • Weighted fusion of classifier scores based on ACKNOWLEDGEMENTS individual classifier performance The authors would like to acknowledge the support from the National Science Foundation under Grant No. 1241585. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation or the US government.
View full slide show




How to work with columns that have default values     Omit columns with default values from the dataset unless they’re needed by the application. Provide values for those columns whenever a row is added to the dataset. Note that although this application will allow data to be added to the Products table, the OnHandQuantity column can be omitted because it’s defined with a default value in the database. So when a new row is added to the database, the database will set this column to its default value. If you include a column with a default value in a dataset, you need to realize hat this value isn’t assigned to the column in the dataset, even though the dataset enforces the constraints for that column. For instance, the OnHandQuantity column in the MMABooks database has a default value of zero and doesn’t allow nulls. But if you include this column in the dataset, its definition will have a default value of null and won’t allow nulls. As a result, an exception will be thrown whenever a new row is added to the dataset with a null value for the OnHandQuantity column. 30
View full slide show




Dataset collection Stan Antol
View full slide show




OCL Quick Reference [http://www.eoinwoods.info/doc/ocl_quick_reference.pdf] Operations are applied to collections using the “->” operator (e.g. items->isEmpty(), where “items” is a collection). Number of times that obj appears in a collection. Collection Manipulation Operations count(obj) excludes(obj) excludesAll(coll) first() includes(obj) includesAll(coll) isEmpty() size() Does count(obj) = 0 ? Does count(obj) = 0 hold for all items in collection coll? The first item in the ordered collection. Is count(obj) > 0 ? Does count(obj) > 0 hold for all items in collection coll? Is collection’s size() = 0 ? Number of items in the collection. Loop Operations collect(expr) Returns a bag containing the value of the expression for each of the items in the collection (e.g. items->collect(value)). A simpler synonym for this operation is the period (“.”) operator (e.g. items.value). forAll(expr) Does expression expr hold for all items in the collection? select(expr) Returns the sub-collection of items in a collection for which expression expr holds. set1->select(attr1 > 10) These two examples are equivalent. set1->select(i | i.attr1 > 10) “i” is an “iterator” variable and can be thought of as being set to each of the elements of set1 in turn. 20
View full slide show




A small part of the hierarchy: class Class Diagram E «interface» lang::Iterable + Collection «interface» util::List iterator() : Iterator Iterable «interface» util::Collection ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ size() : int isEmpty() : boolean contains(Object) : boolean iterator() : Iterator toArray() : Object[] toArray(T[]) : T[] add(E) : boolean remove(Object) : boolean containsAll(Collection) : boolean addAll(Collection) : boolean removeAll(Collection) : boolean retainAll(Collection) : boolean clear() : void equals(Object) : boolean hashCode() : int ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ size() : int isEmpty() : boolean contains(Object) : boolean iterator() : Iterator toArray() : Object[] toArray(T[]) : T[] add(E) : boolean remove(Object) : boolean containsAll(Collection) : boolean addAll(Collection) : boolean addAll(int, Collection) : boolean removeAll(Collection) : boolean retainAll(Collection) : boolean clear() : void equals(Object) : boolean hashCode() : int get(int) : E set(int, E) : E add(int, E) : void remove(int) : E indexOf(Object) : int lastIndexOf(Object) : int listIterator() : ListIterator listIterator(int) : ListIterator subList(int, int) : List CS-2851 Dr. Mark L. Hornick AbstractCollection List util::AbstractList E AbstractList util:: AbstractSequentialList E E AbstractList List Cloneable java.io.Serializable util::ArrayList AbstractSequentialList List Queue Cloneable java.io.Serializable util::LinkedList 3
View full slide show




Strain Distribution in Collections US Collections / BRCs American Type Culture Collection (ATCC) USDA ARS Collection (NRRL) European Collections Deutsche Sammlung vor Microoransmen (DSMZ) Culture Collection University Gottenberg (CCUG) Pasteur Institute (CIP) Laboratory for Micrbiology, Gent (LMG) National Collection of Industrial and Marine Bacteria French Collection of Phytopathogens (CFPB) National Collection of Type Cultures (NCTC) National Collection of Phytopathogenic Bacteria Asia Japan Collection of Microorganisms (JCM) Institute of Fermentation, Osaka (IFO) Korean Collection of Type Cultures (KCTC) 28 Institute of Applied Microbiology, Tokyo (IAM) National Institute of Technology And Evaluation (NBRC) All-Russian Collection of Microorganisms (VKM) Strains 4027 223 1302 183 170 101 25 15 12 11 185 34 26 24 13
View full slide show




Collection is the base interface that many other interfaces, abstract classes, and classes inherit The Collection interface defines certain basic behaviors, such as (to name just a few): add(e) – allows an element to be added to the collection  clear() – removes all elements from the collection  contains(e) – determines if an element is in the collection  isEmpty() – determines if the collection is empty  remove(e) – removes a specific element from the collection  size() – gets the number of elements in the collection Note: Collection does not define a way to retrieve an element from a collection  CS-2851 Dr. Mark L. Hornick 5
View full slide show




Research Contribution  The datasets are in the form of a CSV(Comma separated values) file  The Raw dataset contains the whole Modbus frame    The ARFF dataset contains deep packet inspection of the Modbus frame   Third-party validation of preprocessed ARFF dataset Allows researchers to use specialized preprocessing techniques To be used with WEKA Previous dataset combined four network transactions into one line of the dataset  Each line in the new dataset represents one network transaction
View full slide show




Experiments • Domain • Political Blogosphere • Dataset from Buzzmetrics[2] provides post-post link structure over 14 million posts • Few off-the-topic posts help aggregation • Potential business value • Reference Dataset • Hand-labeled dataset from Lada Adamic et al[3] classifying political blogs into right and left leaning bloggers • Timeframe : 2004 presidential elections, over 1500 blogs analyzed • Overlap of 300 blogs between Buzzmetrics and reference dataset • Goal • Classify the blogs in Buzzmetrics dataset as democrat and republic and compare with reference dataset [2] Lada A. Adamic and Natalie Glance, "The political blogosphere and the 2004 US Election", in Proceedings of the WWW-2005 Workshop Buzzmetrics – www.buzzmetrics.com 03/21/19 Page 33
View full slide show




Experiments Domain • • • • Political Blogosphere Dataset from Buzzmetrics[2] provides post-post links over 1.5M posts Few off-the-topic posts help aggregation Potential business value Reference Dataset • Adamic’s [3] Hand-labeled dataset classifies blogs as right or left leaning • Timeframe: 2004 presidential elections, over 1500 blogs analyzed • Overlap of 300 blogs between Buzzmetrics and reference dataset Goal • Classify the blogs in Buzzmetrics dataset as democrat and republic and compare with reference dataset [2] Lada A. Adamic and Natalie Glance, "The political blogosphere and the 2004 US Election", in Proceedings of the WWW-2005 Workshop, MAY 2005. Buzzmetrics – www.buzzmetrics.com 03/21/19 Page 23
View full slide show




Methodology: Baseline Models: • Support Vector Machine (sk.learn LinearSVC) Accuracy for small dataset(200 images , 2 values of C used , c=.01,1): 50.1% for both c Accuracy for large dataset(141500 images, c=.01): 50% • Random Forest (sklearn.ensemble.RandomForestClassifier ):  100 trees and depth=6 Accuracy for small dataset(200 images): 70% Accuracy for large dataset(141500 images): 70.9% Note: For both SVM and RF , for large dataset we train on 141500 samples, validate on 1000 samples. For small dataset we train on 190 samples , validate on 10 images
View full slide show