Andy Song

RMIT University | RMIT · School of Computer Science and Information Technology

AI Generated Content (AIGC) is becoming phenomenally prominent and impactful. One of the key generative algorithms used in AIGC is the diffusion model which is widely used in generative images and audio. In comparison with other generative methods such as GAN (Generative Adversarial Network) and VAE (Variational Auto Encoder), diffusion models can...

Neural architecture search (NAS) aims to automate architecture engineering in neural networks. This often requires a high computational overhead to evaluate a number of candidate networks from the set of all possible networks in the search space during the search. Prediction of the networks' performance can alleviate this high computational overhea...

Generative adversarial network (GAN) is a powerful method to reproduce the distribution of a given data set. It is widely used for generating photo-realistic images or data collections that appear real. Evolutionary GAN (E-GAN) is one of state-of-the-art GAN variations. E-GAN combines population based search and evolutionary operators from evolutio...

Neural architecture search (NAS) aims to automate architecture engineering in neural networks. This often requires a high computational overhead to evaluate a number of candidate networks from the set of all possible networks in the search space. Prediction of the performance of a network can alleviate this high computational overhead by mitigating...

Existing parking recommendation solutions mainly focus on finding and suggesting parking spaces based on the unoccupied options only. However, there are other factors associated with parking spaces that can influence someone's choice of parking such as fare, parking rule, walking distance to destination, travel time, likelihood to be unoccupied at...

The existing research on parking availability sensing mainly relies on extensive contextual and historical information. In practice, it is challenging to have such information available as it requires continuous collection of sensory signals. In this paper, we design an end-to-end transfer learning framework for parking availability sensing to pred...

Local search algorithms have been successfully used for many combinatorial optimisation problems. The choice of the most suitable local search algorithm is, however, a challenging task as their performance is highly dependent on the problem characteristic. In addition, most of these algorithms require users to select appropriate internal neighbourh...

Finding the shortest route between a pair of origin and destination is known to be a crucial and challenging task in intelligent transportation systems. Current methods assume fixed travel time between any pairs, thus the efficiency of these approaches is limited because the travel time in reality can dynamically change due to factors including the...

The heterogeneous fleet vehicle routing problem with two-dimensional loading constraints (2L- HFVRP) is a complex variant of the classical vehicle routing problem. 2L-HFVRP seeks for minimal cost set of routes to serve a set of customers using a fleet of vehicles of different capacities, fixed and variable operating costs, different dimensions, and...

The authors present an interactive design framework for grammar evolution. A novel aesthetic-based fitness measure was introduced as guidance procedure. One feature of this guidance procedure is the initial input of a reference image. This image provides direction to the evolutionary design process by application of a similarity measure. A case stu...

Deep Belief Networks (DBN) have become a powerful tools to deal with a wide range of applications. On complex tasks like image reconstruction, DBN’s performance is highly sensitive to parameter settings. Manually trying out different parameters is tedious and time consuming however often required in practice as there are not many better options. Th...

http://ecs.victoria.ac.nz/Events/AI2018/WebHome

Deep learning object detectors achieve state-of-the-art accuracy at the expense of high computational overheads, impeding their utilization on embedded systems such as drones. A primary source of these overheads is the exhaustive classification of typically 10^4-10^5 regions per image. Given that most of these regions contain uninformative backgrou...

Deep learning object detectors achieve state-of-the-art accuracy at the expense of high computational overheads, impeding their utilization on embedded systems such as drones. A primary source of these overheads is the exhaustive classification of typically 10^4 to 10^5 regions per image. Given that most of these regions contain uninformative backg...

Image processing and recognition are an important part of the modern society, with applications in fields such as advanced artificial intelligence, smart assistants, and security surveillance. The essential first step involved in almost all the visual tasks is background subtraction with a static camera. Ensuring that this critical step is performe...

The Dynamic Vehicle Routing Problem (DVRP) is a complex variation of classical Vehicle Routing Problem (VRP). The aim of DVRP is to find a set of routes to serve multiple customers at minimal total travelling cost while the travelling time between point to point may vary during the process because of factors like traffic congestion. To effectively...

Visual salience detection originated over 500 million years ago and is one of nature’s most efficient mechanisms. In contrast, many state-of-the-art computational saliency models are complex and inefficient. Most saliency models process high-resolution color images; however, insights into the evolutionary origins of visual salience detection sugges...

Swarm robots are highly desirable in dealing with complex tasks. However, manual coding of individual robot behaviours and robot collaboration is not trivial especially under unknown and dynamic environments. This study introduced a hyper-heuristic methodology for this challenge, so robots can learn suitable behaviours during the process. The hyper...

Project Scheduling Problem (PSP) plays a crucial role in large-scale software development, directly affecting the productivity of the team and on-time delivery of software projects. PSP concerns with the decision of who does what and when during the software project lifetime. PSP is a combinatorial optimisation problem and inherently NP-hard, indic...

Portfolio Selection (PS) is recognized as one of the most important and challenging problems in financial engineering. The aim of PS is to distribute a given amount of investment fund across a set of assets in such a way that the return is maximised and the risk is minimised. To solve PS more effectively and more efficiently, this paper introduces...

Portfolio Selection (PS) is recognized as one of the most important and challenging problems in financial engineering. The aim of PS is to distribute a given amount of investment fund across a set of assets in such a way that the return is maximised and the risk is minimised. To solve PS more effectively and more efficiently, this paper introduces...

A robot swarm is a solution for difficult and large scale tasks. However, controlling and coordinating a swarm of robots is challenging, because of the complexity and uncertainty of the environment where manual programming of robot behaviours is often impractical. In this study we propose a hyper-heuristic methodology for swarm robots. It allows ro...

Deep Convolutional Neural Network (CNN), which is one of the prominent deep learning methods, has shown a remarkable success in a variety of computer vision tasks, especially image classification. However, tuning CNN hyper-parameters requires expert knowledge and a large amount of manual effort of trial and error. In this work, we present the use o...

This paper investigates the Google machine reassignment problem (GMRP). GMRP is a real world optimisation problem which is to maximise the usage of cloud machines. Since GMRP is computationally challenging problem and exact methods are only advisable for small instances, meta-heuristic algorithms have been used to address medium and large instances...

On behalf of the organizing committee, it is our great pleasure to invite you to the annual IEEE Congress on Evolutionary Computation (CEC), which is one of the leading events in the area of evolutionary computation. IEEE CEC provides a forum to bring together researchers and practitioners from all over the world to present and discuss their resear...

Cyber security in the context of big data is known to be a critical problem and presents a great challenge to the research community. Machine learning algorithms have been suggested as candidates for handling big data security problems. Among these algorithms, support vector machines (SVMs) have achieved remarkable success on various classification...

Visual salience detection originated over 500 million years ago and is one of nature's most efficient mechanisms. In contrast, many state-of-the-art computational saliency models are complex and inefficient. Most saliency models process high-resolution color (HC) images; however, insights into the evolutionary origins of visual salience detection s...

This paper, proposes a web-based voting system, which allows voters to cast and submit their electronic ballots by ranking all candidates according to their personal preference. Each ballot is treated as a square matrix, with each element encrypted using the ElGamal cryptosystem before submission. Furthermore, proof of partial knowledge and zero kn...

Google Machine Reassignment Problem (GMRP) is a recent real world problem proposed at ROADEF/EURO challenge 2012. The aim of this problem is to maximise the usage of the available machines by reassigning processes among those machines while a numerous constraints must be not violated. In this work, we propose a great deluge algorithm with multi-nei...

Iterated Local Search (ILS) is a simple yet powerful optimisation method that iteratively invokes a local search procedure with renewed starting points by perturbation. Due to the complexity of search landscape, different ILS strategies may better suit different problem instances or different search stages. To address this issue, this work proposes...

This position paper proposes and defines the nature of a framework, which explores ways of integrating control system (CS) with machine intelligence for generative design (GD). This paper elaborates about the implications of and the potential for impact on GD. The framework described in this work can be used as an active tool to drive design proces...

This paper proposes a automated shape generation methodology based on grammatical genetic programming for specific design cases. Two cases of the shape generation are presented: architectural envelope design and facade design. Through the described experiments, the applicability of this evolutionary method for design applications is showcased. Thro...

It is known that neighbourhood structures affect search performance. In this study we analyse a series of neighbourhood structures to facilitate the search. The well known steepest descent (SD) local search algorithm is used in this study as it is parameter free. The search problem used is the Google Machine Reassignment Problem (GMRP). GMRP is a r...

In this paper, a novel bi-level grouping optimization model is proposed for solving Storage Location Assignment Problem with Grouping Constraint (SLAP-GC). A major challenge of this problem is the grouping constraint which restricts the number of groups each product can have and the locations of items in the same group. In SLAP-GC, the problem cons...

Abstract. Google Machine Reassignment Problem (GMRP) is an optimisation problem proposed at ROADEF/EURO challenge 2012. The task of GMRP is to allocate cloud computing resources by reassigning a set of services to a set of machines while not violating any constraints. We propose an evolutionary parallel late acceptance hill-climbing algorithm (P-LA...

Abstract. Google Machine Reassignment Problem (GMRP) is a real world problem proposed at ROADEF/EURO challenge 2012 competition which must be solved within 5 minutes. GMRP consists in reassigning a set of services into a set of machines for which the aim is to improve the machine usage while satisfying numerous constraints. This paper proposes an e...

Google Machine Reassignment Problem (GMRP) is a real world problem proposed at ROADEF/EURO challenge 2012 competition which must be solved within 5 min. GMRP consists in reassigning a set of services into a set of machines for which the aim is to improve the machine usage while satisfying numerous constraints. This paper proposes an evolutionary si...

A state in time series is time series data stream maintaining a certain pattern over a period of time, for example, holding a steady value, being above a certain threshold and oscillating regularly. Automatic learning and discovery of these patterns of time series states can be useful in a range of scenarios of monitoring and classifying stream dat...

This study investigates the dynamic shortest path routing (DSPR) problem in mobile ad-hoc networks. The goal is to find the shortest possible path that connects a source node with the destination node while effectively handling dynamic changes occurring on the ad-hoc networks. The key challenge in DSPR is how to simultaneously keep track changes an...

We propose a model for clustering data with spatiotemporal intervals, which is a type of spatiotemporal data associated with a start-and an end-point. This model can be used to effectively evaluate clusters of spatiotemporal interval data, which signifies an event at a particular location that stretches over a period of time. Our work aims to deal...

Load balancing (LB) is crucial in the field of cloud computing. LB is to find the optimum allocation of services onto a set of machines so the machine usage can be maximised. This paper proposes a new method for LB, simulated annealing (SA) enhanced by grammatical evolution (GE). SA is a well-known stochastic optimisation algorithm that has good pe...

Load balancing (LB) is an important and challenging optimisation problem in cloud computing. LB involves assigning a set of services into a set of machines for which the goal is to optimise machine usages. This study presents a memetic algorithm (MA) for the LB problem. MA is a hybrid method that combines the strength of population based evolutiona...

This study proposes a novel grammar guided Genetic Programming method to solve a real world problem, the Storage Location Assignment Problem (SLAP) with Grouping Constraints. Self-adaptive Tabu Search algorithms are evolved by this approach and it can be used as solvers for SLAPs. A novel self-adaptive Tabu Search framework is proposed that key con...

The Storage Location Assignment Problem (SLAP) is a significant optimisation problem in warehouse management. Given a number of products, each with a set of items with different popularities (probabilities of being ordered), SLAP is to find the best locations for the items of the products in the warehouse to minimise the warehouse operational cost....

In recent years, wireless sensor networks have been widely used in healthcare applications, such as hospital and home patient monitoring. Wireless medical sensor networks are more vulnerable to eavesdropping, modification, impersonation and replaying attacks than the wired networks. A lot of work has been done to secure wireless medical sensor netw...

The Storage Location Assignment Problem (SLAP) is to find an optimal stock arrangement in a warehouse. This study presents a scalable method for solving large-scale SLAPs utilizing Genetic Programming (GP) and two sampling strategies. Given a large scale problem, a sub-problem is sampled from the original problem for our GP method to learn an alloc...

Anomaly detection aims to find patterns in data that are significantly different from what is defined as normal. One of the challenges of anomaly detection is the lack of labelled examples, especially for the anomalous classes. We describe a neural network based approach to detect anomalous instances using only examples of the normal class in train...

Portfolio Selection (PS) is to allocate a given amount of investment fund across a set of assets in such a way that the return is maximized and the risk is minimized. PS is a challenging financial engineering problem and optimization problem. GA is well known for its effectiveness in solving optimization problems. However it may experience slow con...

Unsafe driving behaviours can put the driver himself and other people participating in the traffic at risk. Smart-phones with built-in inertial sensors offer a convenient way to passively monitor the driving patterns, from which potentially risky events can be detected. However, it is not trivial to decide which sensor data channel is relevant for...

Elderly people are prone to fall due to the high rate of risk factors associated with ageing. Existing fall detection sys-tems are mostly designed for a constrained environment, where various assumptions are applied. To overcome these drawbacks, we opt to use mobile phones with standard built-in sensors. Fall detection is performed on motion data c...

This study proposes a method for solving real-world warehouse Storage Location Assignment Problem (SLAP) under grouping constraints by Genetic Programming (GP). Integer Linear Programming (ILP) formulation is used to define the problem. By the proposed GP method, a subset of the items is repeatedly selected and placed into the available current bes...

Activity recognition from smartphone sensor inputs is of great importance to enhance user experience. Our study aims to investigate the applicability of Genetic Programming (GP) approach on this complex real world problem. Traditional methods often require substantial human efforts to define good features. Moreover the optimal features for one type...

Risky driver behaviours such as sudden braking, swerving, and excessive acceleration are a major risk to road safety. In this study, we present a learning method to recognize such behaviours from smart-phone sensor input which can be considered as a type of multi-channel time series. Unlike other learning methods, this Genetic Programming (GP) base...

Risky driver behaviours such as sudden braking, swerving, and excessive acceleration are a major risk to road safety. In this study, we present a learning method to recognize such behaviours from smartphone sensor input which can be considered as a type of multi-channel time series. Unlike other learning methods, this Genetic Programming (GP) based...

Automated optical inspection (AOI) is desirable in printed circuit board (PCB) manufacturing as inspecting manually is time-consuming and error-prone. This paper presents a study on evolving an AOI program with Genenetic-Programming (GP), an evolution-inspired technique. Using a GP-based approach, domain knowledge such as board design and lighting...

In this paper, we propose a novel hybrid classifica-tion method which is based on two distinct approaches, namely Genetic Programming (GP) and Nearest Neighbour (kNN). The method relies on a memory list which contains some correctly labelled instances and is formed by classifiers evolved by GP. The class label of a new instance will be determined b...

Recognition of activities such as sitting, standing, walking and running can significantly improve the interaction between human and machine, especially on mobile devices. In this study we present a GP based method which can automatically evolve recognition programs for various activities using multisensor data. This investigation shows that GP is...

Recognizing text captured in a photograph, or scene text, remains an unsolved problem in computer vision. Conventional methods require a complex multi-step process to incorporate a pipeline of manually constructed algorithms. In contrast this research presents a single step framework which is based on Genetic Programming (GP). With a suitable metho...

This paper describes an approach to the detection rice plants in images of rice fields by using genetic programming. The method involves the evolution of a genetic programming classifier of 20 × 20 pixel windows to distinguish rice and nonrice windows, applies the evolved classifier to each pixel position in a test image in a scanning window fashio...

A novel approach is proposed in this study, which is to evolve visual inspection programs for automatic defect detection on populated printed circuit boards. This GP-based method does not require knowledge of the layout design of a board, nor relevant domain knowledge such as lighting conditions and visual characteristics of the components. Further...

Compared to conventional activity recognition methods using feature extraction followed by classification, the Genetic Programming (GP) based classification applied to raw sensor data can avoid the time-consuming and knowledge-dependent feature extraction procedure. However, the traditional GP-based classifier using accuracy as fitness function is...

This paper presents an approach to recognition of human actions such as sitting, standing, walking or running by analysing the data produced by the sensors of a smart phone. The data comes as streams of parallel time series from 21 sensors. We have used genetic programming to evolve detectors for a number of actions and compared the detection accur...

Evolving solutions for machine vision applications has gained more popularity in the recent years. One area is evolving programs by Genetic Programming (GP) for motion detection, which is a fundamental component of most vision systems. Despite the good performance, this approach is not widely accepted by mainstream vision application developers. On...

Detecting events of interest in sensor data is crucial in many areas such as medical monitoring by body sensors. Current methods often require prior domain knowledge to be available. Moreover, it is difficult for them to find complex temporal patterns existing in multi-channel data. To overcome these drawbacks, we propose a Genetic Programming (GP)...

Classifying images is of great importance in machine vision and image analysis applications such as object recognition and face detection. Conventional methods build classifiers based on certain types of image features instead of raw pixels because the dimensionality of raw inputs is often too large. Determining an optimal set of features for a par...

This study presents a selective motion detection methodology which is based on genetic programming GP, an evolutionary search strategy. By this approach, motion detection programs can be automatically evolved instead of manually coded. This study investigates the suitable GP representation for motion detection as well as explores the advantages of...

Image classification is a complex but important task especially in the areas of machine vision and image analysis such as remote sensing and face recognition. One of the challenges in image classification is finding an optimal set of features for a particular task because the choice of features has direct impact on the classification performance. H...

Loop structure is a fundamental flow control in programming languages for repeating certain operations. It is not widely used in Genetic Programming as it introduces extra complexity in the search. However in some circumstances, including a loop structure may enable GP to find better solutions. This study investigates the benefits of loop structure...

The aim of event detection in time series is to identify particular occurrences of user-interest in one or more time lines, such as finding an anomaly in electrocardiograms or reporting a sudden variation of voltage in a power supply. Current methods are not adequate for detecting certain kinds of events without any domain knowledge. Therefore, we...

This paper extends the application of Genetic Programming into a new area, automatically splitting video frames based on the content. A GP methodology is presented to show how to evolve a program which can analyse the difference between scenes and split them accordingly. The evolved video splitting programs achieve reasonable performance even when...

Genetic Programming (GP) is reputable for its power in finding creative solutions for complex problems. However the downside of it is also well known: the evolved solutions are often difficult to understand. This interpretability issue hinders GP to gain acceptance from many application areas. To address this issue in the context of motion detectio...

Finding patterns such as increasing or decreasing trends, abrupt changes and periodically repeating sequences is a necessary task in many real world situations. We have shown how genetic programming can be used to detect increasingly complex patterns in time series data. Most classification methods require a hand-crafted feature extraction preproce...