by Neural
Network
- The document proposes a hybrid multi-objective evolutionary algorithm that uses an artificial neural network to reduce the number of objective function evaluations needed. It combines a multi-objective evolutionary algorithm (MOEA) with an artificial neural network (ANN) to approximate solutions. The ANN is trained on solutions evaluated by the MOEA and then used to estimate fitness for unevaluated solutions to further guide the search. This approach aims to improve optimization efficiency over existing MOEAs for problems with computationally expensive objective functions.
SA is a global optimization technique.
It distinguishes between different local optima.
It is a memory less algorithm & the algorithm does not use any information gathered during the search.
SA is motivated by an analogy to annealing in solids.
& it is an iterative improvement algorithm.
This document discusses genetic algorithms and their components. It begins by explaining that genetic algorithms are a type of evolutionary algorithm inspired by biological evolution that uses techniques like inheritance, mutation, selection, and crossover. It then defines the key terms used in genetic algorithms, such as individuals, populations, chromosomes, genes, and fitness functions. The rest of the document provides more details on genetic algorithm components like representation of solutions, selection of individuals, crossover and mutation operations, and the general genetic algorithm process.
Genetic algorithms and traditional algorithms differ in their definitions, usages, and complexity. Genetic algorithms are based on genetics and natural selection, and help find optimal solutions to difficult problems. They are more advanced than traditional algorithms which provide step-by-step procedures. Genetic algorithms are used in fields like machine learning and artificial intelligence, while traditional algorithms are used in programming and mathematics.
This document provides an overview of dynamic programming. It begins by explaining that dynamic programming is a technique for solving optimization problems by breaking them down into overlapping subproblems and storing the results of solved subproblems in a table to avoid recomputing them. It then provides examples of problems that can be solved using dynamic programming, including Fibonacci numbers, binomial coefficients, shortest paths, and optimal binary search trees. The key aspects of dynamic programming algorithms, including defining subproblems and combining their solutions, are also outlined.
Recurrent Neural Networks are popular Deep Learning models that have shown great promise to achieve state-of-the-art results in many tasks like Computer Vision, NLP, Finance and much more. Although being models proposed several years ago, RNN have gained popularity recently. In this talk, we will review how these models evolved over the years, dissection of RNN, current applications and its future.
This document provides an overview of a lecture on designing and analyzing computer algorithms. It discusses key concepts like what an algorithm and program are, common algorithm design techniques like divide-and-conquer and greedy methods, and how to analyze algorithms' time and space complexity. The goals of analyzing algorithms are to understand their behavior, improve efficiency, and determine whether problems can be solved within a reasonable time frame.
SA is a global optimization technique.
It distinguishes between different local optima.
It is a memory less algorithm & the algorithm does not use any information gathered during the search.
SA is motivated by an analogy to annealing in solids.
& it is an iterative improvement algorithm.
This document discusses genetic algorithms and their components. It begins by explaining that genetic algorithms are a type of evolutionary algorithm inspired by biological evolution that uses techniques like inheritance, mutation, selection, and crossover. It then defines the key terms used in genetic algorithms, such as individuals, populations, chromosomes, genes, and fitness functions. The rest of the document provides more details on genetic algorithm components like representation of solutions, selection of individuals, crossover and mutation operations, and the general genetic algorithm process.
Genetic algorithms and traditional algorithms differ in their definitions, usages, and complexity. Genetic algorithms are based on genetics and natural selection, and help find optimal solutions to difficult problems. They are more advanced than traditional algorithms which provide step-by-step procedures. Genetic algorithms are used in fields like machine learning and artificial intelligence, while traditional algorithms are used in programming and mathematics.
This document provides an overview of dynamic programming. It begins by explaining that dynamic programming is a technique for solving optimization problems by breaking them down into overlapping subproblems and storing the results of solved subproblems in a table to avoid recomputing them. It then provides examples of problems that can be solved using dynamic programming, including Fibonacci numbers, binomial coefficients, shortest paths, and optimal binary search trees. The key aspects of dynamic programming algorithms, including defining subproblems and combining their solutions, are also outlined.
Recurrent Neural Networks are popular Deep Learning models that have shown great promise to achieve state-of-the-art results in many tasks like Computer Vision, NLP, Finance and much more. Although being models proposed several years ago, RNN have gained popularity recently. In this talk, we will review how these models evolved over the years, dissection of RNN, current applications and its future.
This document provides an overview of a lecture on designing and analyzing computer algorithms. It discusses key concepts like what an algorithm and program are, common algorithm design techniques like divide-and-conquer and greedy methods, and how to analyze algorithms' time and space complexity. The goals of analyzing algorithms are to understand their behavior, improve efficiency, and determine whether problems can be solved within a reasonable time frame.
The document discusses simulated annealing, a metaheuristic technique inspired by the physical process of annealing in materials. It describes how simulated annealing works by generating random neighbor solutions at each iteration and probabilistically accepting worse solutions based on temperature to avoid local optima. The key concepts are explained, including the Boltzmann distribution used to determine acceptance probability. Parameters like initial temperature, cooling schedule, and stopping criteria are also covered. An example job scheduling problem demonstrates the algorithm. Finally, common applications of simulated annealing in areas like scheduling, routing, and optimization are listed.
The document discusses key concepts in machine learning theory such as sample complexity, computational complexity, and mistake bounds. It focuses on analyzing the performance of broad classes of learning algorithms characterized by their hypothesis space. Specific topics covered include probably approximately correct (PAC) learning, sample complexity for finite vs infinite hypothesis spaces, and mistake bounds for algorithms like HALVING and weighted majority. The goal is to understand how many training examples and computational steps are needed for a learner to converge to a successful hypothesis.
This document discusses algorithms and their analysis. It defines an algorithm as a step-by-step procedure to solve a problem or calculate a quantity. Algorithm analysis involves evaluating memory usage and time complexity. Asymptotics, such as Big-O notation, are used to formalize the growth rates of algorithms. Common sorting algorithms like insertion sort and quicksort are analyzed using recurrence relations to determine their time complexities as O(n^2) and O(nlogn), respectively.
Machine learning and its applications was a gentle introduction to machine learning presented by Dr. Ganesh Neelakanta Iyer. The presentation covered an introduction to machine learning, different types of machine learning problems including classification, regression, and clustering. It also provided examples of applications of machine learning at companies like Facebook, Google, and McDonald's. The presentation concluded with discussing the general machine learning framework and steps involved in working with machine learning problems.
Local search algorithms operate by examining the current node and its neighbors. They are suitable for problems where the solution is the goal state itself rather than the path to get there. Hill-climbing and simulated annealing are examples of local search algorithms. Hill-climbing continuously moves to higher value neighbors until a local peak is reached. Simulated annealing also examines random moves and can accept moves to worse states based on probability. Both aim to find an optimal or near-optimal solution but can get stuck in local optima.
Methods of Optimization in Machine LearningKnoldus Inc.
In this session we will discuss about various methods to optimise a machine learning model and, how we can adjust the hyper-parameters to minimise the cost function.
Stochastic gradient descent and its tuningArsalan Qadri
This paper talks about optimization algorithms used for big data applications. We start with explaining the gradient descent algorithms and its limitations. Later we delve into the stochastic gradient descent algorithms and explore methods to improve it it by adjusting learning rates.
Metaheuristic Algorithms: A Critical AnalysisXin-She Yang
The document discusses metaheuristic algorithms and their application to optimization problems. It provides an overview of several nature-inspired algorithms including particle swarm optimization, firefly algorithm, harmony search, and cuckoo search. It describes how these algorithms were inspired by natural phenomena like swarming behavior, flashing fireflies, and bird breeding. The document also discusses applications of these algorithms to engineering design problems like pressure vessel design and gear box design optimization.
This document discusses classical sets and fuzzy sets. It defines classical sets as having distinct elements that are either fully included or excluded from the set. Fuzzy sets allow for gradual membership, with elements having degrees of membership between 0 and 1. Operations like union, intersection, and complement are defined for both classical and fuzzy sets, with fuzzy set operations accounting for degrees of membership. Properties of classical and fuzzy sets and relations are also covered, noting differences like fuzzy sets not following the law of excluded middle.
Dynamic Programming is a method for solving a complex problem by breaking it down into a collection of simpler subproblems, solving each of those subproblems just once, and storing their solutions using a memory-based data structure (array, map,etc).
This document discusses optimization problems and their solutions. It begins by defining optimization problems as seeking to maximize or minimize a quantity given certain limits or constraints. Both deterministic and stochastic models are discussed. Examples of discrete optimization problems include the traveling salesman and shortest path problems. Solution methods mentioned include integer programming, network algorithms, dynamic programming, and approximation algorithms. The document then focuses on convex optimization problems, which can be solved efficiently. It discusses using tools like CVX for solving convex programs and the duality between primal and dual problems. Finally, it presents the collaborative resource allocation algorithm for solving non-convex optimization problems in a suboptimal way.
1) The document discusses various search algorithms including uninformed searches like breadth-first search as well as informed searches using heuristics.
2) It describes greedy best-first search which uses a heuristic function to select the node closest to the goal at each step, and A* search which uses both path cost and heuristic cost to guide the search.
3) Genetic algorithms are introduced as a search technique that generates successors by combining two parent states through crossover and mutation rather than expanding single nodes.
This document provides an overview and introduction to the course "Knowledge Representation & Reasoning" taught by Ms. Jawairya Bukhari. It discusses the aims of developing skills in knowledge representation and reasoning using different representation methods. It outlines prerequisites like artificial intelligence, logic, and programming. Key topics covered include symbolic and non-symbolic knowledge representation methods, types of knowledge, languages for knowledge representation like propositional logic, and what knowledge representation encompasses.
Design and Analysis of Algorithm help to design the algorithms for solving different types of problems in Computer Science. It also helps to design and analyze the logic of how the program will work before developing the actual code for a program.
Machine Learning With Logistic RegressionKnoldus Inc.
Machine learning is the subfield of computer science that gives computers the ability to learn without being programmed. Logistic Regression is a type of classification algorithm, based on linear regression to evaluate output and to minimize the error.
This document is a preface to a laboratory manual for a Digital Communication Systems course. It discusses the importance of practical, hands-on learning to develop industry-relevant skills. Each practical exercise is designed to help students achieve predetermined outcomes through developing procedures and safety precautions. The course aims to enable students to apply concepts of digital communication to troubleshoot and maintain real-world systems. While efforts were made to eliminate errors, feedback is welcome to improve the manual.
This document describes a multi-objective evolutionary algorithm that uses artificial neural networks to approximate fitness functions in order to reduce the number of exact function evaluations. The algorithm runs the evolutionary algorithm for an initial number of generations to collect a training dataset. It then trains a neural network on this dataset. The evolutionary algorithm continues running for additional generations, using the neural network to approximate some or all of the fitness function evaluations. The neural network approximation error is monitored, and the evolutionary algorithm switches back to using exact function evaluations when the error becomes too high. This process repeats until an acceptable Pareto front is found. The method was tested on benchmark multi-objective test functions and showed a 20-40% reduction in the number of exact function evaluations needed
The document discusses simulated annealing, a metaheuristic technique inspired by the physical process of annealing in materials. It describes how simulated annealing works by generating random neighbor solutions at each iteration and probabilistically accepting worse solutions based on temperature to avoid local optima. The key concepts are explained, including the Boltzmann distribution used to determine acceptance probability. Parameters like initial temperature, cooling schedule, and stopping criteria are also covered. An example job scheduling problem demonstrates the algorithm. Finally, common applications of simulated annealing in areas like scheduling, routing, and optimization are listed.
The document discusses key concepts in machine learning theory such as sample complexity, computational complexity, and mistake bounds. It focuses on analyzing the performance of broad classes of learning algorithms characterized by their hypothesis space. Specific topics covered include probably approximately correct (PAC) learning, sample complexity for finite vs infinite hypothesis spaces, and mistake bounds for algorithms like HALVING and weighted majority. The goal is to understand how many training examples and computational steps are needed for a learner to converge to a successful hypothesis.
This document discusses algorithms and their analysis. It defines an algorithm as a step-by-step procedure to solve a problem or calculate a quantity. Algorithm analysis involves evaluating memory usage and time complexity. Asymptotics, such as Big-O notation, are used to formalize the growth rates of algorithms. Common sorting algorithms like insertion sort and quicksort are analyzed using recurrence relations to determine their time complexities as O(n^2) and O(nlogn), respectively.
Machine learning and its applications was a gentle introduction to machine learning presented by Dr. Ganesh Neelakanta Iyer. The presentation covered an introduction to machine learning, different types of machine learning problems including classification, regression, and clustering. It also provided examples of applications of machine learning at companies like Facebook, Google, and McDonald's. The presentation concluded with discussing the general machine learning framework and steps involved in working with machine learning problems.
Local search algorithms operate by examining the current node and its neighbors. They are suitable for problems where the solution is the goal state itself rather than the path to get there. Hill-climbing and simulated annealing are examples of local search algorithms. Hill-climbing continuously moves to higher value neighbors until a local peak is reached. Simulated annealing also examines random moves and can accept moves to worse states based on probability. Both aim to find an optimal or near-optimal solution but can get stuck in local optima.
Methods of Optimization in Machine LearningKnoldus Inc.
In this session we will discuss about various methods to optimise a machine learning model and, how we can adjust the hyper-parameters to minimise the cost function.
Stochastic gradient descent and its tuningArsalan Qadri
This paper talks about optimization algorithms used for big data applications. We start with explaining the gradient descent algorithms and its limitations. Later we delve into the stochastic gradient descent algorithms and explore methods to improve it it by adjusting learning rates.
Metaheuristic Algorithms: A Critical AnalysisXin-She Yang
The document discusses metaheuristic algorithms and their application to optimization problems. It provides an overview of several nature-inspired algorithms including particle swarm optimization, firefly algorithm, harmony search, and cuckoo search. It describes how these algorithms were inspired by natural phenomena like swarming behavior, flashing fireflies, and bird breeding. The document also discusses applications of these algorithms to engineering design problems like pressure vessel design and gear box design optimization.
This document discusses classical sets and fuzzy sets. It defines classical sets as having distinct elements that are either fully included or excluded from the set. Fuzzy sets allow for gradual membership, with elements having degrees of membership between 0 and 1. Operations like union, intersection, and complement are defined for both classical and fuzzy sets, with fuzzy set operations accounting for degrees of membership. Properties of classical and fuzzy sets and relations are also covered, noting differences like fuzzy sets not following the law of excluded middle.
Dynamic Programming is a method for solving a complex problem by breaking it down into a collection of simpler subproblems, solving each of those subproblems just once, and storing their solutions using a memory-based data structure (array, map,etc).
This document discusses optimization problems and their solutions. It begins by defining optimization problems as seeking to maximize or minimize a quantity given certain limits or constraints. Both deterministic and stochastic models are discussed. Examples of discrete optimization problems include the traveling salesman and shortest path problems. Solution methods mentioned include integer programming, network algorithms, dynamic programming, and approximation algorithms. The document then focuses on convex optimization problems, which can be solved efficiently. It discusses using tools like CVX for solving convex programs and the duality between primal and dual problems. Finally, it presents the collaborative resource allocation algorithm for solving non-convex optimization problems in a suboptimal way.
1) The document discusses various search algorithms including uninformed searches like breadth-first search as well as informed searches using heuristics.
2) It describes greedy best-first search which uses a heuristic function to select the node closest to the goal at each step, and A* search which uses both path cost and heuristic cost to guide the search.
3) Genetic algorithms are introduced as a search technique that generates successors by combining two parent states through crossover and mutation rather than expanding single nodes.
This document provides an overview and introduction to the course "Knowledge Representation & Reasoning" taught by Ms. Jawairya Bukhari. It discusses the aims of developing skills in knowledge representation and reasoning using different representation methods. It outlines prerequisites like artificial intelligence, logic, and programming. Key topics covered include symbolic and non-symbolic knowledge representation methods, types of knowledge, languages for knowledge representation like propositional logic, and what knowledge representation encompasses.
Design and Analysis of Algorithm help to design the algorithms for solving different types of problems in Computer Science. It also helps to design and analyze the logic of how the program will work before developing the actual code for a program.
Machine Learning With Logistic RegressionKnoldus Inc.
Machine learning is the subfield of computer science that gives computers the ability to learn without being programmed. Logistic Regression is a type of classification algorithm, based on linear regression to evaluate output and to minimize the error.
This document is a preface to a laboratory manual for a Digital Communication Systems course. It discusses the importance of practical, hands-on learning to develop industry-relevant skills. Each practical exercise is designed to help students achieve predetermined outcomes through developing procedures and safety precautions. The course aims to enable students to apply concepts of digital communication to troubleshoot and maintain real-world systems. While efforts were made to eliminate errors, feedback is welcome to improve the manual.
This document describes a multi-objective evolutionary algorithm that uses artificial neural networks to approximate fitness functions in order to reduce the number of exact function evaluations. The algorithm runs the evolutionary algorithm for an initial number of generations to collect a training dataset. It then trains a neural network on this dataset. The evolutionary algorithm continues running for additional generations, using the neural network to approximate some or all of the fitness function evaluations. The neural network approximation error is monitored, and the evolutionary algorithm switches back to using exact function evaluations when the error becomes too high. This process repeats until an acceptable Pareto front is found. The method was tested on benchmark multi-objective test functions and showed a 20-40% reduction in the number of exact function evaluations needed
This document summarizes a presentation on modified Booth multipliers and FIR filters. It introduces Booth multiplication as an efficient algorithm for signed number multiplication. It then discusses FIR filters and their structure, noting that they require many multiplications. The Radix-2 and Radix-4 Booth multiplication algorithms are described in steps. Simulation results demonstrating the Radix-4 algorithm multiplying two numbers are shown. References on digital filter design and implementation are provided at the end.
Definition and Validation of Scientific Algorithms for the SEOSAT/Ingenio GPPEsri
Presentation by Eduardo de Miguel, Raúl Valenzuela, Teodoro Bernardino, Verena Rodríguez, Alberto Pizarro, Diana de Miguel and Severino Fernández from INTA, GMV and EADS-CASA made on Esri European User Conference 2011.
Convolutional Neural Network Architecture and Input Volume Matrix Design for ...Takumi Kodama
This document describes a study that used convolutional neural networks (CNNs) to improve classification accuracy in a tactile P300-based brain-computer interface (BCI). The researchers designed CNN input volumes from EEG signal intervals and implemented a LeNet-inspired CNN architecture. Classification accuracy improved from around 80% with raw EEG data to 100% when using moving averages. The CNN approach allowed for non-personalized training, demonstrating the potential of CNNs to advance tactile P300 BCI accuracy and usability.
Crude-Oil Scheduling Technology: moving from simulation to optimizationBrenno Menezes
Scheduling technology either commercial or homegrown in today’s crude-oil refining industries relies on a complex simulation of scenarios where the user is solely responsible for making many different decisions manually in the search for feasible solutions over some limited time-horizon i.e., trial-and-error heuristics. As a normal outcome, schedulers abandon these solutions and then return to their simpler spreadsheet simulators due to: (i) time-consuming efforts to configure and manage numerous scheduling scenarios, and (ii) requirements of updating premises and situations that are constantly changing. Moving to solutions based in optimization rather than simulation, the lecture describes the future steps in the refactoring of the scheduling technology in PETROBRAS considering in separate the graphic user interface (GUI) and data communication developments (non-modeling related), and the modeling and process engineering related in an automated decision-making with built-in problem representation facilities and integrated data handling features among other techniques in a smart scheduling frontline.
HPC on Cloud for SMEs. The case of bolt tightening.Andrés Gómez
This document discusses using high performance computing (HPC) resources in the cloud to help small and medium enterprises (SMEs) perform simulations. It describes a case study where HPC resources were used to simulate the bolt tightening process for an SME called Texas Controls. The simulations used Code_Aster software to model the materials, design, sequence and tightening parameters. A Taguchi method was employed to automatically generate 16 parametric simulation jobs. Results were analyzed to determine the optimal tightening strategy. Remote visualization and a graphical user interface were provided to make the HPC resources accessible to the SME. The model was also validated against real sensor data to verify accuracy.
Analysis of Educational Robotics activities using a machine learning approachLorenzo Cesaretti
These slides present the preliminary results through the utilisation of machine learning techniques for the analysis of Educational Robotics activities. An experimentation with 197 secondary school students from Italy was con-ducted, through updating Lego Mindstorms EV3 programming blocks in order to record log files containing the coding sequences designed by the students (within team work), during the resolution of a preliminary Robotics’ exercise. We utilised four machine learning techniques (logistic regression, support vec-tor machine, K-nearest neighbors and random forests) to predict the students’ performance, comparing a supervised approach (using twelve indicators ex-tracted from the log files as input for the algorithms) and a mixed approach (ap-plying a k-means algorithm to calculate the machine learning features). The re-sults have highlighted that SVM with the mixed approach outperformed the other techniques, and that three learning styles were predominantly emerged from the data mining analysis.
Full Body Spatial Vibrotactile Brain Computer Interface ParadigmTakumi Kodama
This document summarizes research on a full-body spatial vibrotactile brain-computer interface (BCI) paradigm. The research aims to 1) develop a new touch-based BCI intended for communicating with ALS patients and 2) confirm the effectiveness of the modality by improving stimulus pattern classification accuracies. The approach involves applying six vibrotactile stimulus patterns to a user's back while they are lying down. A series of experiments were conducted including psychophysical testing, online EEG classification, and offline classification refinement using machine learning algorithms like SVM and CNN. The results confirmed the validity and feasibility of the full-body tactile BCI paradigm, achieving up to 100% classification accuracy using a CNN model trained on data
This document outlines the scheme and syllabus for an undergraduate program in physical sciences (physics, computer science, and mathematics) under the Choice Based Credit System at Guru Jambheshwar University of Science and Technology, Hisar. The program consists of 120 credits over six semesters, including foundation courses (72 credits), discipline-specific electives (36 credits), and skill enhancement courses (6 credits). Courses include language skills, awareness programs, core courses, practical labs, and seminars. The syllabus and credit distribution for individual courses are provided, along with the internal and external assessment scheme.
This document outlines the regulations, curriculum, and syllabi for the M.E. Computer Science and Engineering program following a choice-based credit system at Anna University, Chennai. It provides the program educational objectives, outcomes, mapping of objectives to outcomes, course details over 4 semesters including theory and practical courses. The document also lists the foundation, professional core, professional elective, and employability enhancement courses along with their course codes, titles, categories and credit details. Core areas of study include advanced algorithms, software engineering, operating systems, databases, security and elective courses on topics such as data mining, networks, and machine learning.
Syllabus for fourth year of engineeringtakshakpdesai
The document discusses revisions to the Bachelor of Engineering Computer Engineering program at the University of Mumbai. Key points include:
1. The curriculum is being revised to incorporate outcome-based education and a semester-based credit and grading system to improve quality and ensure excellence in engineering education.
2. Program educational objectives and course objectives/outcomes are being clearly defined to support outcome-based learning.
3. Revisions include new/updated courses in the 7th and 8th semesters, such as Digital Signal Processing, Cryptography, and Data Warehousing and Mining.
4. The credit and grading system is being implemented progressively starting with the 1st year of the program through to the final
2014 04 03 (educon2014) emadrid uned a practice based mooc for learning elect...eMadrid network
This document discusses the implementation of a MOOC for learning electronics using a remote laboratory called VISIR. The MOOC was the world's first remote lab-based MOOC. It consisted of 8 modules over 10 hours each that introduced circuit simulation tools and then had students do real experiments using VISIR. Over its first two editions, the MOOC had a diverse set of students from different backgrounds and countries. The experiments using VISIR allowed up to 384 simultaneous students to experiment remotely. Future work aims to expand the types of circuits and experiments available through the MOOC and VISIR platform.
M.tech.(cse)(regular) part ii(semester iii & iv)1Rekha Bhatia
This document outlines the syllabus for the third and fourth semesters of the M.Tech Computer Science and Engineering program at Punjabi University, Patiala for the 2017-2018 academic session.
In the third semester, students will take courses in Network Security, Neural Networks and Fuzzy Logic, Digital Image Processing, a research project, and a software lab or elective course. Assessment will include internal exams, assignments, attendance, and participation.
The fourth semester will consist solely of a dissertation worth 21 credits and 400 marks.
INTERNSHIP TRAINING CHENNAI-IPT FOR CSE/IT/ECE/E&I-MAASTECHASHOKKUMAR RAMAR
The document discusses an inplant training program conducted by MAASTECH for students in various engineering fields such as ECE, EEE, CSE, IT, and Biomedical. The training covers various topics including basic electronics, embedded systems, PIC microcontroller programming, sensor interfacing, circuit design software, and involves hands-on projects. Students who attend will receive a certificate and training lasts 30-20 hours depending on the field. MAASTECH is located in Chennai and provides hostel facilities for trainees.
INPLANT TRAINING IN CHENNAI FOR ECE STUDENTS,EEE STUDENTSASHOKKUMAR RAMAR
The document discusses an inplant training program conducted by MAASTECH for students in various engineering fields such as ECE, EEE, CSE, IT, and Biomedical. The training covers various topics including basic electronics, embedded systems, PIC microcontroller programming, sensor interfacing, circuit design software, and involves hands-on projects. Students who attend will receive a certificate and training lasts 30-20 hours depending on the field. MAASTECH is located in Chennai and provides hostel facilities nearby for students attending the program.
INPLANT TRAINING FOR 3RD YEAR STUDENTS-ECE/EEE/E&I/ICE/ETE-MAASTECHASHOKKUMAR RAMAR
The document discusses an inplant training program conducted by MAASTECH for students in various engineering fields such as ECE, EEE, CSE, IT, and Biomedical. The training includes courses in basic electronics, embedded systems, PCB assembly, sensor interfacing, programming with C and VB6.0. Students work on sample projects and get involved in ongoing R&D products. Training is provided in circuit design, robotics, GSM and other applications. A fee of Rs. 1000 is charged per student for the training period, after which an IPT certificate will be issued. Hostel facilities are also available nearby.
INPLANT TRAINING-FUNDAMENTAL OF ELECTRONICS/EMBEDDED SYSTEMSASHOKKUMAR RAMAR
The document discusses an inplant training program conducted by MAASTECH for students in fields like CSE, IT, ECE, EEE, biomedical, and more. It provides details of the training program contents which include basics of electronics, embedded systems, PIC microcontroller programming, sensor interfacing, circuit design software, and involvement in ongoing projects. It specifies the fees as Rs. 1000 per student and requirements like minimum batch size, bonafide certificate, and training certificate awarded. Contact information and details of hostel facilities are also provided.
Improving Insurance Risk Prediction with Generative Adversarial Networks (GANs)Armando Vieira
Generative adversarial networks (GANs) show promise for addressing data imbalance issues in insurance modeling. GANs were originally developed for computer vision tasks but have also been applied to tabular data. Conditional GANs and CycleGANs can generate synthetic minority class examples to balance datasets. In a case study on insurance fraud detection, GANs outperformed traditional resampling techniques like SMOTE in improving precision, recall, and F1-score. However, GANs require dense feature representations and consistency over time to be effective for tabular data imbalance problems.
Predicting online user behaviour using deep learning algorithmsArmando Vieira
We propose a robust classifier to predict buying intentions based on user behaviour within a large e-commerce website. In this work we compare traditional machine learning techniques with the most advanced deep learning approaches. We show that both Deep Belief Networks and Stacked Denoising auto-Encoders achieved a substantial improvement by extracting features from high dimensional data during the pre-train phase. They prove also to be more convenient to deal with severe class imbalance.
Boosting conversion rates on ecommerce using deep learning algorithmsArmando Vieira
This document summarizes an approach to use deep learning algorithms to predict the probability that online shoppers will purchase a product based on their website interactions. The approach involves using stacked auto-encoders to reduce the high dimensionality of the product interaction data before applying classification algorithms. Testing on various datasets showed that random forest outperformed logistic regression and that incorporating time data and more training examples improved prediction performance. Further work proposed applying stacked auto-encoders and deep belief networks to fully leverage the large amount of product interaction data.
Seasonality effects on second hand cars salesArmando Vieira
This document analyzes seasonality effects on car sales using weekly aggregated car deal data from October 2012 to November 2014. It finds that:
1) A sudden drop in the last week's sales can be explained by statistical fluctuations based on the normal distribution of weekly deals over the period.
2) Months with the lowest deals (November and December) still show that last week's sales of 154 were a normal occurrence based on the mean and standard deviation for those months.
3) Google trends data for the keyword "used cars" shows a clear seasonality pattern of decreasing searches before the end of the year and increasing searches at the start and middle of the year.
Visualizations of high dimensional data using R and ShinyArmando Vieira
This document discusses building interactive visualizations with Shiny and R to explore social and health care data from the UK. It describes using inputs like demographics, economic deprivation, and health metrics to create outputs like a health score and stress score. Visualizations were created with Shiny and Google Motion Charts to compare districts. The document concludes discussing using machine learning techniques like embeddings and exploring causality.
The document discusses GPU computing for machine learning. It notes that machine learning algorithms are computationally expensive and their requirements increase with data size. GPUs provide significant performance gains over CPUs for parallel problems like machine learning. Many machine learning algorithms have been implemented on GPUs, achieving speedups of 1-2 orders of magnitude. However, most GPU implementations are closed-source. Open-source implementations provide advantages like reproducibility and fair algorithm comparisons.
This document provides an overview of deep learning algorithms, including deep neural networks, convolutional neural networks, deep belief networks, and restricted Boltzmann machines. It discusses key concepts such as learning in deep neural networks, the evolution timeline of deep learning approaches, deep architectures, and restricted Boltzmann machines. It also covers training restricted Boltzmann machines using contrastive divergence, constructing deep belief networks by stacking restricted Boltzmann machines, and practical considerations for pre-training and fine-tuning deep belief networks.
Extracting Knowledge from Pydata London 2015Armando Vieira
The document discusses using deep learning techniques like word embeddings to jointly embed text and knowledge graphs for information extraction purposes. Word embeddings represent words as vectors in a way that captures semantic meaning, allowing related words to have similar embeddings. Knowledge graphs explicitly represent entities and relations. The document proposes combining text corpora with knowledge graphs by training a model on both to generate embeddings that incorporate information from both sources. This allows extracting knowledge expressed in text and transforming it into a machine-readable format.
We propose an algorithm for training Multi Layer Preceptrons for classification problems, that we named Hidden Layer Learning Vector Quantization (H-LVQ). It consists of applying Learning Vector Quantization to the last hidden layer of a MLP and it gave very successful results on problems containing a large number of correlated inputs. It was applied with excellent results on classification of Rurtherford
backscattering spectra and on a benchmark problem of image recognition. It may also be used for efficient feature extraction.
machine learning in the age of big data: new approaches and business applicat...Armando Vieira
Presentation at University of Lisbon on Machine Learning and big data.
Deep learning algorithms and applications to credit risk analysis, churn detection and recommendation algorithms
Optimization of digital marketing campaignsArmando Vieira
This document discusses using machine learning techniques to optimize digital marketing campaigns. Specifically, it analyzes data from campaigns using clustering, visualization and predictive models. Unsupervised learning methods like k-means clustering, PCA, MDS and SOM are used to identify patterns in large digital data. Supervised models like SVMs and random forests predict conversions. The goal is to extract actionable insights to improve ROI, engagement and sales through optimization of parameters like ad design, keywords, bids, channels and budget allocation.
Credit risk with neural networks bankruptcy prediction machine learningArmando Vieira
The document discusses credit risk management with AI tools. It summarizes that credit scoring is used to statistically quantify risk by converting applicant information into numbers and a score. The objective is to forecast future performance based on past client behavior. It then discusses using various machine learning models like HLVQ-C and neural networks to predict financial distress, classify companies, and improve bankruptcy prediction. The models were tested on real world credit and financial datasets.
This document outlines a proposal called "Democracy 2" which aims to define a new democratic model that is more citizen-centric and suited to today's society. It proposes moving beyond representative democracy by giving citizens a more direct role in important political decisions through information technology. The initiative will define the new model through contributions from citizens and experts across three streams focusing on political, social, and technology issues. It will also conduct a proof of concept trial of the new model at the local/regional level in multiple countries. The overall goal is to create a more open and representative democratic system.
Sairmais.com is a new tourism web portal that uses a recommendation system to provide personalized recommendations to users. It analyzes a user's social connections and preferences to filter vast amounts of tourism information and provide the most relevant options. The portal aims to be a one-stop platform for comprehensive geo-referenced tourism data. It incorporates review sharing and social networking features commonly seen on sites like Amazon, Facebook and TripAdvisor. Sairmais.com's recommendation system analyzes the relationships between users, items, and ratings to provide customized recommendations tailored to each individual user's interests. The system seeks to simplify the travel planning process and provide a more personal touch than other major tourism websites.
Sairmais.com is a new tourism web portal that uses a recommendation system to provide personalized recommendations to users. It analyzes a user's social connections and ratings of tourism items like hotels and restaurants to filter vast amounts of online tourism information and provide the most relevant options. The portal aims to be a one-stop platform for comprehensive geo-referenced tourism data. It incorporates social networking features allowing users to share experiences and opinions to improve recommendations for others. The system utilizes collaborative tagging and ratings within a user's social network to build profiles and predict their preferences, helping users more easily plan trips by finding the best options tailored specifically for them.
Manifold learning for bankruptcy predictionArmando Vieira
This document presents a method for bankruptcy prediction and analysis using manifold learning. Specifically, it applies the Isomap algorithm with class label information incorporated into the dissimilarity matrix (S-Isomap) on a real dataset of French companies. S-Isomap is shown to have comparable testing accuracy to other classifiers like SVM and better than KNN and RVM, while providing excellent lower-dimensional visualization with only 3 dimensions. The S-Isomap approach achieves separability of patterns from healthy to bankrupt firms in the embedded space. This preprocessing technique using manifold learning is a promising approach for bankruptcy prediction and analysis on high-dimensional financial data.
This document presents a study comparing several machine learning models for personal credit scoring: logistic regression, multilayer perceptron, support vector machine, AdaBoostM1, and Hidden Layer Learning Vector Quantization (HLVQ-C). The models were tested on datasets from a Portuguese bank. HLVQ-C achieved the highest accuracy and was the most useful model according to a proposed measure that considers earnings from denying bad credits and losses from denying good credits. While other models had higher error rates for good credits, HLVQ-C balanced accuracy and usefulness the best, making it suitable for commercial credit scoring applications.
O autor descreve como a curiosidade natural das crianças é inibida pelo sistema educativo, transformando o ensino da ciência em algo abstrato e livresco em vez de prático e exploratório. Isto leva ao desinteresse dos alunos pela ciência e ao pequeno papel de Portugal na investigação científica. Defende que a educação deve estimular a curiosidade das crianças em vez de a reprimir.
Artificial neural networks for ion beam analysisArmando Vieira
The document discusses using artificial neural networks (ANNs) for ion beam analysis. Specifically, it discusses:
1) Using ANNs to analyze Rutherford backscattering spectroscopy (RBS) data in an automated way, by recognizing patterns in the data related to sample properties without explicit knowledge of causes.
2) Training ANNs on datasets of RBS spectra with known sample parameters to allow the ANNs to relate spectral features to things like layer thickness, composition, and depth.
3) The potential for ANNs to enable real-time automated analysis and optimization of ion beam experiments.
The document discusses artificial intelligence and pattern recognition. It introduces various pattern recognition concepts including defining a pattern, examples of patterns in different domains, and approaches to pattern recognition. It also provides an example of using discriminative methods to classify fish into salmon and sea bass using optical sensing and extracted features.
THE SACRIFICE HOW PRO-PALESTINE PROTESTS STUDENTS ARE SACRIFICING TO CHANGE T...indexPub
The recent surge in pro-Palestine student activism has prompted significant responses from universities, ranging from negotiations and divestment commitments to increased transparency about investments in companies supporting the war on Gaza. This activism has led to the cessation of student encampments but also highlighted the substantial sacrifices made by students, including academic disruptions and personal risks. The primary drivers of these protests are poor university administration, lack of transparency, and inadequate communication between officials and students. This study examines the profound emotional, psychological, and professional impacts on students engaged in pro-Palestine protests, focusing on Generation Z's (Gen-Z) activism dynamics. This paper explores the significant sacrifices made by these students and even the professors supporting the pro-Palestine movement, with a focus on recent global movements. Through an in-depth analysis of printed and electronic media, the study examines the impacts of these sacrifices on the academic and personal lives of those involved. The paper highlights examples from various universities, demonstrating student activism's long-term and short-term effects, including disciplinary actions, social backlash, and career implications. The researchers also explore the broader implications of student sacrifices. The findings reveal that these sacrifices are driven by a profound commitment to justice and human rights, and are influenced by the increasing availability of information, peer interactions, and personal convictions. The study also discusses the broader implications of this activism, comparing it to historical precedents and assessing its potential to influence policy and public opinion. The emotional and psychological toll on student activists is significant, but their sense of purpose and community support mitigates some of these challenges. However, the researchers call for acknowledging the broader Impact of these sacrifices on the future global movement of FreePalestine.
Gender and Mental Health - Counselling and Family Therapy Applications and In...PsychoTech Services
A proprietary approach developed by bringing together the best of learning theories from Psychology, design principles from the world of visualization, and pedagogical methods from over a decade of training experience, that enables you to: Learn better, faster!
Andreas Schleicher presents PISA 2022 Volume III - Creative Thinking - 18 Jun...EduSkills OECD
Andreas Schleicher, Director of Education and Skills at the OECD presents at the launch of PISA 2022 Volume III - Creative Minds, Creative Schools on 18 June 2024.
Level 3 NCEA - NZ: A Nation In the Making 1872 - 1900 SML.pptHenry Hollis
The History of NZ 1870-1900.
Making of a Nation.
From the NZ Wars to Liberals,
Richard Seddon, George Grey,
Social Laboratory, New Zealand,
Confiscations, Kotahitanga, Kingitanga, Parliament, Suffrage, Repudiation, Economic Change, Agriculture, Gold Mining, Timber, Flax, Sheep, Dairying,
How Barcodes Can Be Leveraged Within Odoo 17Celine George
In this presentation, we will explore how barcodes can be leveraged within Odoo 17 to streamline our manufacturing processes. We will cover the configuration steps, how to utilize barcodes in different manufacturing scenarios, and the overall benefits of implementing this technology.
Neural Networks and Genetic Algorithms Multiobjective acceleration
1. 1
A Hybrid Multi-Objective Evolutionary Algorithm
Using an Inverse Neural Network
A. Gaspar-Cunha(1), A. Vieira(2), C.M. Fonseca(3)
(1)
IPC- Institute for Polymers and Composites, Dept. of Polymer Engineering,
University of Minho, Guimarães, Portugal
(2)
ISEP and Computational Physics Centre,
Coimbra, Portugal
(3)
CSI- Centre for Intelligent Systems, Faculty of Science and Technology,
University of Algarve, Faro, Portugal
HYBRID METAHEURISTICS (HM 2004)
ECAI 2004, Valencia, Spain
August, 2004
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
2. 2
INTRODUCTION
Most real optimization problems are multiobjective
Example: Simultaneous minimization of the cost and maximization
of the performance of a specific system
Dominated solution
Cost
Single optimum
(maximal performance)
Performance
Single optimum
(minimal cost)
Instituto Superior de
Engenharia do Porto
Multiple optima
(both objectives optimized)
PARETO FRONTIER
(set of non-dominated solutions)
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
3. 3
INTRODUCTION
Computation time required to evaluate the solutions
Start
Engineering problems:
Initialise Population
i=0
Black Box
Numerical modelling
routines
• Finite elements
• Finite differences
• Finite volumes
• etc
Evaluation
Assign Fitness
Fi
Convergence
criterion
satisfied?
i=i+1
no
Selection
HIGH COMPUTATION TIMES
yes
Stop
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Recombination
Dept. Polymer Engineering
University of Minho
4. 4
INTRODUCTION
OBJECTIVES:
• Develop an efficient multi-objective optimization
algorithm
• Reduce the number of evaluations of objective
functions necessary
• Compare performance with existing algorithms
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
5. 5
CONTENTS
• Multi-Objective Evolutionary Algorithm (MOEA)
• Artificial Neural Networks (ANN)
• Hybrid Multi-Objective Algorithm (MOEA-IANN)
• Results and Discussion
• Conclusions
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
6. MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM – MOEA6
How to deal with multiple criteria (or objectives)?
Single objective
(for example, weighted sum)
0 ≤ wj ≤ 1
∑ wj = 1
0 ≤ Fj ≤ 1
0 ≤ FOi ≤ 1
q
FOi = ∑ w j F j
j =1
Decision made before the search
Pareto Frontier
Multiobjective optimization
Decision made after the search
Objective 2
200
1
190
180
2
170
5
6
3
4
160
500
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
1000
Objective 1
1500
Dept. Polymer Engineering
University of Minho
2000
7. MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM – MOEA7
Basic functions of a MOEA:
Maintaining a diverse
nondominated set
(Density estimation)
Density
C2
Archiving
Fitness
C1
Instituto Superior de
Engenharia do Porto
Preventing nondominated
solutions from being lost
(Elitist population - archiving)
Guiding the population
towards the Pareto set
(Fitness assignment)
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
8. MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM – MOEA8
Reduced Pareto Set G.A. with Elitism (RPSGAe)
Start
RPSGAe sorts the population individuals in a number of
pre-defined ranks using a clustering technique, in order
to reduce the number of solutions on the efficient
frontier.
Initialise Population
i=0
a) Rank the individuals using a clustering
Evaluation
algorithm;
b) Calculate
Assign Fitness
Fi
i=i+1
the
fitness
using
a
ranking
function;
c) Copy the best individuals to the external
population;
Convergence
criterion
satisfied?
no
Selection
yes
Stop
Instituto Superior de
Engenharia do Porto
Recombination
d) If the external population becomes full:
- Apply the clustering algorithm to the
external population;
- Copy the best individuals to the internal
population;
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
9. MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM – MOEA9
Clustering algorithm example
NR = r
N=15; Nranks=3
N ranks
r=2; NR=10
r=1; NR=5
C2
N
C2
1
1
2
1
12
1
2
1
2
1
1
2
1
C1
1
C1
Gaspar-Cunha, A., Covas, J.A. - RPSGAe - A Multiobjective Genetic Algorithm with Elitism: Application
to Polymer Extrusion, in Metaheuristics for Multiobjective Optimisation, Lecture Notes in Economics and
Mathematical Systems, Gandibleux, X.; Sevaux, M.; Sörensen, K.; T'kindt, V. (Eds.), Springer, 2004.
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
10. MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM – MOEA10
Clustering algorithm example
r=3; NR=15
Fitness - Linear ranking :
C2
1
2( SP − 1) ( N + 1 − i )
FOi = 2 − SP +
N
23
12 3
2
1 3
2
FO(1) = 2.00
FO(2) = 1.87
31
23
FO(3) = 1.73
1
C1
RPSGAe
• Number of Ranks - Nranks
Parameters:
• Limits of indifference of the clustering algorithm - limit
• N. of individuals copied to the external population - Next
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
11. MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM – MOEA11
Reduced Pareto Set G.A. with Elitism (RPSGAe)
Internal
population
Next
Internal
population
(Generation n)
External
population
External
population
(Generation n)
Generation 1
Generation 2
Generation 3
Generation 4
Next
Generation 5
Generation n
Order of the RPSGAe: O(Nranks q N2)
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
12. MULTI-OBJECTIVE EVOLUTIONARY ALGORITHM – MOEA12
How the basic functions are accomplished in the RPSGAe :
1. Guiding the population towards the Pareto set
Fitness assignment: ranking function based
reduction of the Pareto Set
on the
2. Maintaining a diverse nondominated set
Density estimation: ranking function based on the reduction
of the Pareto Set
3. Preventing nondominated solutions from being lost
Elitist population: periodic copy of the best solutions (to the
main population), selected with the method of Pareto set
reduction
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
13. 13
ARTIFICIAL NEURAL NETWORKS – ANN
Artificial Neural Networks
•
ANN implemented by a Multilayer Preceptron is a flexible scheme capable of
approximating an arbitrary complex function;
•
The ANN builds a map between a set of inputs and the respective outputs;
•
A feed-forward neural network consists of an
array of input nodes connected to an array of
output nodes through successive intermediate
layers;
•
•
Each connection between nodes has a weight,
initially random, which is adjusted during a
training process;
The output of each node of a specific layer is a
function of the sum on the weighted signals
coming from the previous layer;
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Input
Layer
Hidden
Layer
Output
Layer
P1
C1
P2
C2
...
...
Pi
Cj
Dept. Polymer Engineering
University of Minho
14. 14
HYBRID MULTI-OBJECTIVE ALGORITHM
Two possible approachs to reduce the computation time
1. During evaluation – Some solutions can be evaluated
using an approximate function, such as Fitness Inheritance,
Artificial Neural Networks, etc (this reduce the number of
exact evaluations necessary).
2. During recombination – Some individuals can be
generated using more efficient methods (this produce a fast
approximation to the optimal Pareto frontier, thus the
number of generations is reduced).
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
15. 15
HYBRID MULTI-OBJECTIVE ALGORITHM – MOEA-ANN
Use of ANN to “Evaluate” some Solutions
Start
Artificial Neural Network
Initialise Population
i=0
Parameters
to optimise
P1
Convergence
criterion
satisfied?
i=i+1
no
P2
C2
...
Pi
Assign Fitness
Fi
C1
...
Evaluation
Criteria
Cj
Selection
yes
Stop
Instituto Superior de
Engenharia do Porto
Recombination
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
16. 16
HYBRID MULTI-OBJECTIVE ALGORITHM – MOEA-ANN
Use of ANN to “Evaluate” some Solutions – Method A
Proposed by K. Deb et. al
Neural Network
learning using
some solutions
of the p
generations
Neural Network
learning using
some solutions
of the p
generations
p generations r generations
p generations r generations
RPSGA with RPSGA with
Neural
exact
Network
function
evaluation
evaluation
RPSGA with RPSGA with
Neural
exact
Network
function
evaluation
evaluation
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
...
...
p generations
RPSGA with
exact
function
evaluation
Dept. Polymer Engineering
University of Minho
17. 17
HYBRID MULTI-OBJECTIVE ALGORITHM – MOEA-ANN
Use of ANN to “Evaluate” some Solutions – Method B
Neural Network
learning using
some solutions
of the p
generations
Neural Network
learning using
some solutions
of the p
generations
eNN > allowed error
p generations r generations
RPSGA with RPSGA with:
exact
• All solutions
function
(N) evaluated
evaluation
by Neural
Network
• M evaluated
by exact
function
Instituto Superior de
Engenharia do Porto
M
e NN =
S
j =1
(C
NN
i, j
i =1
∑ ∑
− Ci , j
)
2
S
M
eNN > allowed error
p generations r generations
RPSGA with RPSGA with:
exact
• All solutions
function
(N) evaluated
evaluation
by Neural
Network
• M evaluated
by exact
function
Faculty of Science and Technology
University of Algarve
...
...
p generations
RPSGA with
exact
function
evaluation
Dept. Polymer Engineering
University of Minho
18. HYBRID MULTI-OBJECTIVE ALGORITHM – MOEA-IANN
18
Use of an Inverse ANN as “Recombination” operator
Start
Recombination operators:
Initialise Population
• Crossover
i=0
• Mutation
• Inverse ANN (IANN)
Evaluation
Criteria
Variables
C1
V1
C2
V2
Selection
...
...
Recombination
Cq
VM
Assign Fitness
Fi
Convergence
criterion
satisfied?
i=i+1
no
yes
Stop
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
19. 19
HYBRID MULTI-OBJECTIVE ALGORITHM – MOEA-IANN
Set of Solutions Generated with the IANN
Selection of n+q solutions from the
• 3.q extreme solutions
• n interior solutions
For j = 1, ..., q
(where, q is the number of criteria) :
∆C2
c
Criterion 2
present population to generate:
b
e1
C j = C 'j + ∆C j
Points 1, 2, …, n:
a
1
2
3
a
4
e2
b
c
Criterion 1 ∆C1
Point ej to a: C j = C j + ∆C j
'
Point ej to b: C j ( j =i ) = C 'j
∧ C j ( j ≠i ) = C 'j + ∆C j
Point ej to c: C j ( j =i ) = C j − ∆C j
'
Instituto Superior de
Engenharia do Porto
∧ C j ( j ≠i ) = C 'j + ∆C j
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
20. HYBRID MULTI-OBJECTIVE ALGORITHM – MOEA-IANN
Set of Solutions Generated with the IANN
Use of IANN to generate
new solutions
c
e1
a
1
2
3
a
4
e2
Parameter 2
Criterion 2
∆C2
b
b
c
Criterion 1 ∆C1
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
2
1
b
a
c
4
e1
a
3
e2
b
c
Parameter 1
Dept. Polymer Engineering
University of Minho
20
21. HYBRID MULTI-OBJECTIVE ALGORITHM – MOEA-IANN
21
MOEA-IANN Algorithm Parameters
Number of Ranks - Nranks
N. of individuals copied to the external population - Next
Limits of indifference of the clustering algorithm – limit
Criteria variation at beginning - ∆Cinit
Criteria variation at end - ∆Cf
N. of generations which individuals are used to train the IANN – Ngen
Rate of individuals generated with the IANN – IR
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
22. 22
RESULTS AND DISCUSSION – Test problems
K. Deb et. al - Test Problem Generator
Minimize f1 ( x1 ) ,
Minimize f 2 ( x2 ) ,
Minimize
f q −1 ( xq −1 ),
f q ( x ) = g ( xq ) h( f1 ( x1 ) , f 2 ( x2 ) , , f q −1 ( xq −1 ), g ( xq ) ),
Minimize
Subject to
x
xi ∈ ℜ i , for i = 1, 2, , q.
2 Criteria
2C-ZDT1 (Convex): M = 30; xi ∈ [0, 1]
1.00
f1 ( x1 ) = x1
= g × 1 −
where, g ( x 2 , , x M
Instituto Superior de
Engenharia do Porto
0.60
f1
g
) = 1+ 9 ∑
M
i =2
f2
f 2 ( x 2 , , x M )
0.80
0.40
0.20
xi
M −1
Faculty of Science and Technology
University of Algarve
0.00
0
0.2
0.4
0.6
0.8
f1
Dept. Polymer Engineering
University of Minho
1
23. 23
RESULTS AND DISCUSSION – Test problems
2 Criteria
2C-ZDT2 (Non-convex): M = 30; xi ∈ [0, 1]
1.00
f1 ( x1 ) = x1
f1 2
= g × 1 −
g
where, g ( x 2 , , x M ) = 1 + 9
∑
M
i=2
0.60
f2
f 2 ( x 2 , , x M )
0.80
0.40
0.20
xi
0.00
M −1
0
0.4
0.6
1
1.00
f1 ( x1 ) = x1
0.60
f1
− f 1 sin (10 π f1 )
g g
) = 1+ 9 ∑
M
0.20
f2
f 2 ( x 2 , , x M ) = g × 1 −
-0.200
xi
0.4
0.6
-0.60
M −1
0.2
-1.00
i=2
f1
Instituto Superior de
Engenharia do Porto
0.8
f1
2C-ZDT3 (Discrete): M = 30; xi ∈ [0, 1]
where, g ( x 2 , , x M
0.2
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
0.8
1
24. 24
RESULTS AND DISCUSSION – Test problems
2 Criteria
2C-ZDT4 (Multimodal): M = 10; x1 ∈ [0, 1]; xi ∈ [-5, 5]
1.40
1.20
f1 ( x1 ) = x1
= g × 1 −
f1
g
f2
f 2 ( x 2 , , x M )
1.00
0.80
0.60
(
where, g ( x 2 , , x M ) = 1 + 10 ( M − 1) + ∑i = 2 xi2 − 10 cos( 4 π xi )
M
0.40
)
0.20
0.00
0
0.2
0.4
0.6
0.8
1
0.6
0.8
1
f1
2C-ZDT6 (Non-uniform): M = 10; xi ∈ [0, 1]
1.00
f1 ( x1 ) = 1 − exp(−4 x1 ) sin 6 (6 π x1 )
f1 2
= g × 1 −
g
where, g ( x 2 , , x M )
Instituto Superior de
Engenharia do Porto
∑M xi
= 1 + 9 i = 2
M −1
0.60
f2
f 2 ( x 2 , , x M )
0.80
0.40
0.25
Faculty of Science and Technology
University of Algarve
0.20
0.00
0
0.2
0.4
f1
Dept. Polymer Engineering
University of Minho
25. 25
RESULTS AND DISCUSSION – Test problems
3 Criteria
3C-ZDT1 (Convex): M = 30; xi ∈ [0, 1]
f1 ( x1 ) = x1
1.0
f 3 ( x 3 , , x M )
= g × 1 −
where, g ( x3 , , x M
f1 f 2
g
) = 1+ 9 ∑
M
i =3
f3
f 2 ( x2 ) = x2
0.5
0.0
0.2
0.4
xi
0.6
f2
M −1
0.4
0.6
0.8
0.8
1.0
0.2
0.0
0.0
f1
1.0
3C-ZDT2 (Non-convex): M = 30; xi ∈ [0, 1]
f1 ( x1 ) = x1
1.0
f 3 ( x3 , , x M )
f f 2
= g × 1 − 1 2
g
where, g ( x3 , , xM
Instituto Superior de
Engenharia do Porto
) =1+ 9 ∑
M
i =3
xi
M −1
Faculty of Science and Technology
University of Algarve
f3
f 2 ( x2 ) = x 2
0.5
0.0
0.2
0.4
0.6
f2
0.4
0.6
0.8
0.8
1.0
1.0
Dept. Polymer Engineering
University of Minho
f1
0.2
0.0
0.0
26. 26
RESULTS AND DISCUSSION – Test problems
3 Criteria
3C-ZDT3 (Discrete): M = 30; xi ∈ [0, 1]
f1 ( x1 ) = x1
1.00
f 2 ( x2 ) = x2
0.75
0.50
f1 f 2 f1 f 2
sin (10 π f1 f 2 )
−
g
g
0.00
0.0
0.2
0.4
0.6
∑i = 3 x i
M
f2
where, g ( x3 , , x M ) = 1 + 9
0.25
f3
f 3 ( x 3 , , x M ) = g × 1 −
M −1
-0.25
0.8
1.0
1.00
1.0
0.8
0.6
0.4
0.2
f1
1.0
0.75
1.000
0.8
0.8125
0.50
0.6250
0.25
0.4375
f3
0.6
0.2500
f2
0.00
0.06250
0.4
-0.1250
-0.25
-0.50
0.0
0.2
0.4
f1 0.6
0.6
0.4
0.8
0.2
1.0
Instituto Superior de
Engenharia do Porto
f2
0.8
1.0
-0.3125
0.2
0.0
0.0
-0.5000
0.2
0.4
0.6
0.8
1.0
f1
0.0
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
0.0
-0.50
27. 27
RESULTS AND DISCUSSION – Test problems
3 Criteria
3C-ZDT4 (Multimodal): M = 10; x1,2 ∈ [0, 1]; xi ∈ [-5, 5]
18
f1 ( x1 ) = x1
16
14
f 2 ( x2 ) = x2
f 3 ( x 3 , , x M )
12
= g × 1 −
f1 f 2
g
f3
10
8
6
4
(
where, g ( x3 , , x M ) = 1 + 10 ( M − 1) + ∑i =3 xi2 − 10 cos( 4 π xi )
M
)
0.0
0.2
0.4
0.6
f2
3C-ZDT6 (Non-uniform): M = 10; xi ∈ [0, 1]
0.4
0.6
0.8
0.8
1.0
0.2
f1
1.0
f 1 ( x1 ) = 1 − exp(−4 x1 ) sin 6 (6 π x1 )
1.0
f 2 ( x 2 ) = 1 − exp(−4 x 2 ) sin 6 (6 π x 2 )
f f 2
= g × 1 − 1 2
g
where, g ( x3 , , x M )
Instituto Superior de
Engenharia do Porto
0.8
∑ xi
= 1 + 9 i =3
M −1
M
0.6
f3
f 3 ( x 3 , , x M )
2
0.0
0
0.4
0.25
Faculty of Science and Technology
University of Algarve
0.2
0.0
0.2
0.4
0.6
f2
0.4
0.6
0.8
0.8
1.0
1.0
Dept. Polymer Engineering
University of Minho
f1
0.2
0.0
0.0
28. 28
RESULTS AND DISCUSSION – Metrics
Hypervolume Metric (Zitzler and Thiele - 1998)
This metric calculates the dominated space volume,
enclosed by the nondominated points and the origin.
S metric:
Volume of the space dominated by
the set of objective vectors
C2
Hypervolume
C1
Criteria C1 and C2 to maximize
Instituto Superior de
Engenharia do Porto
However, is not possible to say
that one set is better than other
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
29. 29
RESULTS AND DISCUSSION – Algorithm Parameters
Influence of algorithm parameters on performance
Parameter
Tested values(*)
Best results
Influence
Selected
limit
0.01; 0.05; 0.1; 0.2
[0.01; 0.2]
Small
0.01
∆ Cinit
0.3; 0.4; 0.5; 0.6
[0.3; 0.5]
Small
0.5
∆ Cf
0.0; 0.1; 0.2; 0.3
[0.0; 0.3]
Small
0.2
Ngen
5; 10; 15; 20
[5; 10]
Small
5
IR
0.35; 0.50; 0.65; 0.80
[0.35; 0.8]
Small
0.8
(*) 5 runs for each tested parameter value
• The influence of the algorithm parameters on its
performance is very small.
• Each optimisation run was carried out 21 times
using the algorithm parameters selected and
different seed values.
Algorithm Parameters:
- N = 100
- Ne = 100
- Nranks = 30
- Next = 3N/Nranks = 10
- cR = 0.8
- mR = 0.05
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
30. 30
RESULTS AND DISCUSSION – Method B
Use of ANN to “Evaluate” some Solutions – Method B
S metric, 22000 evaluations
Number of evaluations
Test
problem
Method B
RPSGAe
Decrease (%)
Method B
RPSGAe
Decrease (%)
ZDT1
0.851
0.849
0.24
10000
19000
47.4
ZDT2
0.786
0.773
1.68
15300
22000
30.5
ZDT3
2.736
2.554
7.13
18000
22000
18.2
ZDT4
0.1116
0.0807
38.29
5000
22000
77.3
ZDT6
0.599
0.571
4.90
12500
22000
43.2
• The S metric after 22000 evaluations decrease when Method B is
used
• The number of evaluations necessary to attain identical level of the
S metric decreases considerably when Method B is used
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
31. RESULTS AND DISCUSSION – 2 Criteria Test Problems
MOEA - Inverse ANN
2C-ZDT1
1
S metric
0.8
0.6
0.4
IANN
RPSGAe
0.2
0
0
50
100
150
Generations
200
250
300
• The Inverse ANN approach has the largest improvement during the
first generations, i.e., when the solution is far from the optimum;
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
31
32. 32
RESULTS AND DISCUSSION – 2 Criteria Test Problems
MOEA - Inverse ANN
2C-ZDT2
2C-ZDT3
2.5
S metric
3
0.8
S metric
1
0.6
0.4
IANN
RPSGAe
0.2
2
1.5
1
0
0
0
100
Generations
200
300
0
2C-ZDT4
0.15
100 Generations 200
300
2C-ZDT6
0.8
0.6
0.1
S metric
S metric
IANN
RPSGAe
0.5
0.05
IANN
RPSGAe
0
0.4
0.2
IANN
RPSGAe
0
0
Instituto Superior de
Engenharia do Porto
100Generations 200
300
Faculty of Science and Technology
University of Algarve
0
100
Generations
200
Dept. Polymer Engineering
University of Minho
300
33. RESULTS AND DISCUSSION – 3 Criteria Test Problems
MOEA - Inverse ANN
3C-ZDT1
0.8
S metric
0.6
0.4
IANN
0.2
RPSGAe
0
0
50
100
150
200
250
300
Generations
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
33
34. 34
RESULTS AND DISCUSSION – 3 Criteria Test Problems
MOEA - Inverse ANN
3C-ZDT2
0.8
1.5
S metric
S metric
0.6
0.4
0.2
IANN
RPSGAe
1.2
0.9
0.6
IANN
RPSGAe
0.3
0
0
0
100 Generations 200
300
0
3C-ZDT4
0.06
100 Generations 200
300
3C-ZDT6
0.4
0.3
0.04
S metric
S metric
3C-ZDT3
1.8
0.02
0.2
0.1
IANN
RPSGAe
0
IANN
RPSGAe
0
0
Instituto Superior de
Engenharia do Porto
100Generations 200
300
Faculty of Science and Technology
University of Algarve
0
100 Generations 200
Dept. Polymer Engineering
University of Minho
300
35. 35
CONCLUSIONS
• Algorithm parameters have a limited influence on its
performance
• Good performance of the proposed algorithm
• The number of generations needed to reach identical level
of performance is reduced thus, the computation time is
reduced by more than 50%.
• Most improvements of the IANN approach
accomplished during the first generations
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho
are
36. 36
ANY QUESTION!?
Instituto Superior de
Engenharia do Porto
Faculty of Science and Technology
University of Algarve
Dept. Polymer Engineering
University of Minho