A comprehensive review of computer-aided whole-slide image analysis: from datasets to feature extraction,segmentation, classification and detection approaches

With the development of Computer-aided Diagnosis (CAD) and image scanning techniques, Whole-slide Image (WSI) scanners are widely used in the field of pathological diagnosis. Therefore, WSI analysis has become the key to modern digital histopathology. Since 2004, WSI has been used widely in CAD. Since machine vision methods are usually based on semi-automatic or fully automatic computer algorithms, they are highly efficient and labor-saving. The combination of WSI and CAD technologies for segmentation, classification, and detection helps histopathologists to obtain more stable and quantitative results with minimum labor costs and improved diagnosis objectivity. This paper reviews the methods of WSI analysis based on machine learning. Firstly, the development status of WSI and CAD methods are introduced. Secondly, we discuss publicly available WSI datasets and evaluation metrics for segmentation, classification, and detection tasks. Then, the latest development of machine learning techniques in WSI segmentation, classification, and detection are reviewed. Finally, the existing methods are studied, and the application prospects of the methods in this field are forecasted.

     

Application of the OBD method for optimization of neural state variable estimators of the two-mass drive system

This paper presents neural estimators of the mechanical state variables of the electrical drive system with elastic joints. The non-measurable state variables, as the torsional torque and the load machine speed are estimated using multilayer feed-forward neural networks. The main stages of the design methodology of these neural estimators are presented. The optimal brain damage method is implemented for the structure optimization of each neural network. Then signals estimated by neural estimators are tested in the electrical drive control structure with additional feedbacks from the estimated shaft torque and the difference between the motor and the load speeds. The simulation results show good accuracy of both presented neural estimators for the wide range of changes of the reference speed and the load torque. The simulation results are then verified by laboratory experiments.     

Is Evolution Algorithmic?

In Darwin’s Dangerous Idea, Daniel Dennett claims that evolution is algorithmic. On Dennett’s analysis, evolutionary processes are trivially algorithmic because he assumes that all natural processes are algorithmic. I will argue that there are more robust ways to understand algorithmic processes that make the claim that evolution is algorithmic empirical and not conceptual. While laws of nature can be seen as compression algorithms of information about the world, it does not follow logically that they are implemented as algorithms by physical processes. For that to be true, the processes have to be part of computational systems. The basic difference between mere simulation and real computing is having proper causal structure. I will show what kind of requirements this poses for natural evolutionary processes if they are to be computational.     

Rigorous Numerical Approach to Isolation in Dynamical Systems on the Example of the Kuramoto-Sivashinsky Equation

This paper describes a method of rigorous verification of an isolating neighborhood based on computer assisted computations. As an application we study the Kuramoto-Sivashinsky equation. The result of the computer assisted proof was directly used in [9] to prove the existence of heteroclinic solutions of the Kuramoto-Sivashinsky equation.     

Training neural networks on high-dimensional data using random projection

Pattern Analysis and Applications - Training deep neural networks (DNNs) on high-dimensional data with no spatial structure poses a major computational problem. It implies a network architecture...     

Detecting Fixed Patterns in Chordal Graphs in Polynomial Time

The Contractibility problem takes as input two graphs G and H, and the task is to decide whether H can be obtained from G by a sequence of edge contractions. The Induced Minor and Induced Topological Minor problems are similar, but the first allows both edge contractions and vertex deletions, whereas the latter allows only vertex deletions and vertex dissolutions. All three problems are NP-complete, even for certain fixed graphs H. We show that these problems can be solved in polynomial time for every fixed H when the input graph G is chordal. Our results can be considered tight, since these problems are known to be W[1]-hard on chordal graphs when parameterized by the size of H. To solve Contractibility and Induced Minor, we define and use a generalization of the classic Disjoint Paths problem, where we require the vertices of each of the k paths to be chosen from a specified set. We prove that this variant is NP-complete even when k=2, but that it is polynomial-time solvable on chordal graphs for every fixed k. Our algorithm for Induced Topological Minor is based on another generalization of Disjoint Paths called Induced Disjoint Paths, where the vertices from different paths may no longer be adjacent. We show that this problem, which is known to be NP-complete when k=2, can be solved in polynomial time on chordal graphs even when k is part of the input. Our results fit into the general framework of graph containment problems, where the aim is to decide whether a graph can be modified into another graph by a sequence of specified graph operations. Allowing combinations of the four well-known operations edge deletion, edge contraction, vertex deletion, and vertex dissolution results in the following ten containment relations: (induced) minor, (induced) topological minor, (induced) subgraph, (induced) spanning subgraph, dissolution, and contraction. Our results, combined with existing results, settle the complexity of each of the ten corresponding containment problems on chordal graphs.     

A fault-tolerant control strategy for non-linear discrete-time systems: application to the twin-rotor system

In this paper, an active fault-tolerant control scheme is proposed in the case of actuator faults. In particular, the general idea of integrating fault identification and control schemes, which takes into account the fault estimation error is first presented in a linear context. As a result, the so-called separation principle for the controller and the fault identification scheme is developed. Subsequently, the proposed approach is extended to a class of non-linear systems. Similarly to the linear case, it is proven that using a suitable control strategy and a faulty identification scheme it is possible to obtain an integrated fault-tolerant control framework, which takes into account the fault identification error. As a result, a non-linear counterpart of the above-mentioned separation principle is developed. Finally, the last part of the paper shows the application results obtained using a twin-rotor system that confirm the high performance of the proposed approach.     

Experiences with large-scale operational trials of ALTO-enhanced P2P filesharing in an intra-ISP scenario

Application Layer Traffic Optimization (ALTO) has recently gained attention in the research and standardisation community as a way for a network operator to guide the peer selection process of distributed applications by providing network layer topology information. In particular P2P applications are expected to gain from ALTO, due to the many connections peers form among each other, often without taking network layer topology information into account. In this paper, we present results of an extensive intra-ISP trial with an ALTO-enhanced P2P filesharing software. In summary, our results show that—depending on the concrete setting and on the distribution of upload capacity in the network—ALTO enables an ISP to save operational costs significantly while not degrading application layer performance noticeably. In addition, based on our experience we are able to give advice to operators on how to save costs with ALTO while not sacrificing application layer performance at all.