Pattern Recognition and Machine Learning

doi:10.1198/TECH.2007.S518

Journal Article•DOI•

Pattern Recognition and Machine Learning

01 Aug 2007-Technometrics (Taylor & Francis)-Vol. 49, Iss: 3, pp 366-366

TL;DR: This book covers a broad range of topics for regular factorial designs and presents all of the material in very mathematical fashion and will surely become an invaluable resource for researchers and graduate students doing research in the design of factorial experiments.

read less

Abstract: (2007). Pattern Recognition and Machine Learning. Technometrics: Vol. 49, No. 3, pp. 366-366.

...read moreread less

Citations

PDF

Open Access

More filters

Journal Article•DOI•

Combining shape, texture and intensity features for cell nuclei extraction in Pap smear images

[...]

Marina E. Plissiti¹, Christophoros Nikou¹, Antonia Charchanti¹•Institutions (1)

University of Ioannina¹

01 Apr 2011-Pattern Recognition Letters

TL;DR: Comparisons with the segmentation results of a gradient vector flow deformable (GVF) model and a region based active contour model (ACM) are performed, which indicate that the proposed method produces more accurate nuclei boundaries that are closer to the ground truth.

...read moreread less

149 citations

Journal Article•DOI•

Analyzing the Training Processes of Deep Generative Models

[...]

Mengchen Liu¹, Jiaxin Shi¹, Kelei Cao¹, Jun Zhu¹, Shixia Liu¹ - Show less +1 more•Institutions (1)

Tsinghua University¹

01 Jan 2018-IEEE Transactions on Visualization and Computer Graphics

TL;DR: A visual analytics approach for better understanding and diagnosing the training process of a deep generative models (DGMs) and proposes a credit assignment algorithm that indicates how other neurons contribute to the output of the neuron causing the training failure.

...read moreread less

Abstract: Among the many types of deep models, deep generative models (DGMs) provide a solution to the important problem of unsupervised and semi-supervised learning. However, training DGMs requires more skill, experience, and know-how because their training is more complex than other types of deep models such as convolutional neural networks (CNNs). We develop a visual analytics approach for better understanding and diagnosing the training process of a DGM. To help experts understand the overall training process, we first extract a large amount of time series data that represents training dynamics (e.g., activation changes over time). A blue-noise polyline sampling scheme is then introduced to select time series samples, which can both preserve outliers and reduce visual clutter. To further investigate the root cause of a failed training process, we propose a credit assignment algorithm that indicates how other neurons contribute to the output of the neuron causing the training failure. Two case studies are conducted with machine learning experts to demonstrate how our approach helps understand and diagnose the training processes of DGMs. We also show how our approach can be directly used to analyze other types of deep models, such as CNNs.

...read moreread less

149 citations

Cites methods from "Pattern Recognition and Machine Lea..."

...For the backward contribution, we leverage the backpropagation algorithm [6], which clearly discloses how the outputs of neurons in layer l+1 indirectly influence the outputs of the neurons in layer l....
[...]
...To address this issue, we first tried to cluster the neurons using the popular K-Means clustering algorithm [6] and only show the clusters whose neurons highly contribute to the output of the selected neurons (Fig....
[...]
...According to the backpropagation algorithm [6], the output al+1 k of the neuron n l+1 k in layer l + 1 has a backward contribution on the gradient gi j of weight wi j....
[...]

Proceedings Article•DOI•

The Door

[...]

Byongsue Kang, Hyo-Houn No, Jung-Hwan Sung¹, Jaeyoung Kim¹, Hwanik Jo¹, Semi Kim¹, Min-Ju Lee - Show less +3 more•Institutions (1)

Soongsil University¹

12 Dec 2011

TL;DR: This work focuses on a door because it links one place to another, and the linking of different places and sharing of places is one of the substantial qualities of network technology.

...read moreread less

Abstract: We want to express digital communication with a human touch. So we pay attention to the properties of a door because it links one place to another. We think the linking of different places and sharing of places is one of the substantial qualities of network technology. Another reason that we focus on a door is because it bears great meanings. For example, a door is the first spot for a meeting with someone or for kissing someone goodbye. It could be your children, spouse, or your friends. These days, we talk and say hello to friends by Messenger or mobiles. Accustomed to digital devices which neglect time or space, we expect that the devices can carry our emotions to others. However, these have limits in expression. We want to show through the our work "The Door" that network technology should have a little more humanity and love in the near future.

...read moreread less

149 citations

Cites background from "Pattern Recognition and Machine Lea..."

...The Bayesian Motivation or a Motivation for Bayes Finally, there exists yet another approach to statistics, called Bayesian inference, which is very popular in, for example, the field of machine learning [Bishop, 2006]....
[...]

Journal Article•DOI•

Effects of user mental state on EEG-BCI performance

[...]

Andrew Myrden¹, Andrew Myrden², Tom Chau², Tom Chau¹•Institutions (2)

University of Toronto¹, Holland Bloorview Kids Rehabilitation Hospital²

02 Jun 2015-Frontiers in Human Neuroscience

TL;DR: It is indicated that mental state is closely related to BCI performance, encouraging future development of psychologically adaptive BCIs and showing the sensitivity of underlying class distributions to changes in mental state.

...read moreread less

Abstract: Changes in psychological state have been proposed as a cause of variation in brain-computer interface performance, but little formal analysis has been conducted to support this hypothesis. In this study, we investigated the effects of three mental states - fatigue, frustration, and attention - on BCI performance. Twelve able-bodied participants were trained to use a two-class EEG-BCI based on the performance of user-specific mental tasks. Following training, participants completed three testing sessions, during which they used the BCI to play a simple maze navigation game while periodically reporting their perceived levels of fatigue, frustration, and attention. Statistical analysis indicated that there is a significant relationship between frustration and BCI performance while the relationship between fatigue and BCI performance approached significance. BCI performance was 7% lower than average when self-reported fatigue was low and 10% lower than average when self-reported frustration was low. A multivariate analysis of mental state revealed the presence of contiguous regions in mental state space where BCI performance was more accurate than average, suggesting the importance of moderate fatigue for achieving effortless focus on BCI control, frustration as a potential motivating factor, and attention as a compensatory mechanism to increasing frustration. Finally, a visual analysis showed the sensitivity of underlying class distributions to changes in mental state. Collectively, these results indicate that mental state is closely related to BCI performance, encouraging future development of psychologically adaptive BCIs.

...read moreread less

149 citations

Cites methods from "Pattern Recognition and Machine Lea..."

...For each feature space dimensionality, a linear discriminant analysis (LDA) classifier (Bishop, 2006) was trained for each candidate task and feature selection method....
[...]
...Each trial was 15 s long, consisting of a 5-s preparation period during which a visual cue was displayed to indicate the required task for the trial; a 5-s task period during which the participant performed the required task; and a 5-s cool-down period before the next trial began....
[...]

Journal Article•DOI•

Multipolicy decision-making for autonomous driving via changepoint-based behavior prediction: Theory and experiment

[...]

Enric Galceran¹, Alexander G. Cunningham², Ryan M. Eustice³, Edwin Olson³•Institutions (3)

Institute of Robotics and Intelligent Systems¹, Toyota², University of Michigan³

09 Feb 2017-Autonomous Robots

TL;DR: This paper reports on an integrated inference and decision-making approach for autonomous driving that models vehicle behavior for both the authors' vehicle and nearby vehicles as a discrete set of closed-loop policies.

...read moreread less

Abstract: This paper reports on an integrated inference and decision-making approach for autonomous driving that models vehicle behavior for both our vehicle and nearby vehicles as a discrete set of closed-loop policies. Each policy captures a distinct high-level behavior and intention, such as driving along a lane or turning at an intersection. We first employ Bayesian changepoint detection on the observed history of nearby cars to estimate the distribution over potential policies that each nearby car might be executing. We then sample policy assignments from these distributions to obtain high-likelihood actions for each participating vehicle, and perform closed-loop forward simulation to predict the outcome for each sampled policy assignment. After evaluating these predicted outcomes, we execute the policy with the maximum expected reward value. We validate behavioral prediction and decision-making using simulated and real-world experiments.

...read moreread less

149 citations

Cites methods from "Pattern Recognition and Machine Lea..."

...Given a segment from time s to t and a policy π, CHAMP approximates the logarithm of the policy evidence for that segment via the Bayesian information criterion (BIC) [4] as:...
[...]
...The BIC is a well-known approximation that avoids marginalizing over the policy parameters and provides a principled penalty against complex policies by assuming a Gaussian posterior around the estimated parameters θ̂....
[...]
...Given a segment from time s to t and a policy π, CHAMP approximates the logarithm of the policy evidence for that segment via the Bayesian information criterion (BIC) [4] as: logL(s, t, π) ≈ log p(zs+1:t|π, θ̂)− 1 2 kπ log(t− s), (9) where kπ is the number of parameters of policy π and θ̂ are estimated parameters for policy π....
[...]

Collapse

Pattern Recognition and Machine Learning

Citations

Cites methods from "Pattern Recognition and Machine Lea..."

Cites background from "Pattern Recognition and Machine Lea..."

Cites methods from "Pattern Recognition and Machine Lea..."

Cites methods from "Pattern Recognition and Machine Lea..."

Related Papers (5)