Cluster sampling | EPFL Graph Search

In statistics, cluster sampling is a sampling plan used when mutually homogeneous yet internally heterogeneous groupings are evident in a statistical population. It is often used in marketing research. In this sampling plan, the total population is divided into these groups (known as clusters) and a simple random sample of the groups is selected. The elements in each cluster are then sampled. If all elements in each sampled cluster are sampled, then this is referred to as a "one-stage" cluster sampling plan. If a simple random subsample of elements is selected within each of these groups, this is referred to as a "two-stage" cluster sampling plan. A common motivation for cluster sampling is to reduce the total number of interviews and costs given the desired accuracy. For a fixed sample size, the expected random error is smaller when most of the variation in the population is present internally within the groups, and not between the groups. Cluster elemental The populati

Multimodal person recognition in audio-visual streams

Do Hoang Nam Le

Multimedia databases are growing rapidly in size in the digital age. To increase the value of these data and to enhance the user experience, there is a need to make these videos searchable through automatic indexing. Because people appearing and talking in the videos are often of high interest for end users, indices that represent the location and identity of people in the archive are indispensable for video search and browsing tools. On the other hand, multimedia videos contain resourceful data of people in both visual and auditory domains. This offers a potential for multimodal learning in the task of human identification. Hence, the main theme of this thesis is on algorithms to create indexes and exploit the audio-visual correspondence in large multimedia corpuses based on person identities. First, this thesis deals with algorithms to create indexes through person discovery in videos. It involves several components: face and speaker diarization, face-voice association, and person naming. To obtain face clusters, we propose a novel face tracking approach that leverages face detectors with a tracking-by-detection framework relying on long term time-interval sensitive association costs. We use also shot context to further accelerate and improve face clustering. Face clusters are then associated to speaker clusters using dubbing and talking detection, in which a multimodal framework is introduced to represent the temporal relationship between the auditory and visual streams. We also improve speaker embeddings for recognition and clustering by using a regularizer called intra-class loss. In the second half, the thesis focuses on multimodal learning with face-voice data. Here, we aim to answer two research questions. First, can one improve a voice embedding using knowledge transferred from a face representation? We investigate several transfer learning approaches to constrain the target voice embedding space to share latent attributes with the source face embedding space. The crossmodal constrains act as regularizers helping voice models, especially in the low-data setting. The second question is can face clusters be used as training labels to learn a speaker embedding? To answer this, we explore the tolerance of embedding losses under label uncertainty. From the risk minimization perspective, we obtain the analytical results that provide the heuristics in strategies to improve the tolerance against label noise. We apply the findings into our task of learning speaker embeddings using face clusters as labels. While the experimental results agree with the analytical heuristics, there is still a large gap in performance between the supervised and the weakly supervised models, which requires further investigation in the future.

EPFL2019

Systematic analysis of low-affinity transcription factor binding site clusters in vitro and in vivo establishes their functional relevance

Shiyu Cheng, Ivan Istomin, Maria Lopez Malo, Sebastian Maerkl, Evan James Olson, Amir Shahein

Binding to binding site clusters has yet to be characterized in depth, and the functional relevance of low-affinity clusters remains uncertain. We characterized transcription factor binding to low-affinity clusters in vitro and found that transcription factors can bind concurrently to overlapping sites, challenging the notion of binding exclusivity. Furthermore, small clusters with binding sites an order of magnitude lower in affinity give rise to high mean occupancies at physiologically-relevant transcription factor concentrations. To assess whether the observed in vitro occupancies translate to transcriptional activation in vivo, we tested low-affinity binding site clusters in a synthetic and native gene regulatory network in S. cerevisiae. In both systems, clusters of low-affinity binding sites generated transcriptional output comparable to single or even multiple consensus sites. This systematic characterization demonstrates that clusters of low-affinity binding sites achieve substantial occupancies, and that this occupancy can drive expression in eukaryotic promoters.

NATURE PORTFOLIO2022

Experimental Study on Bedload Transport and Bedforms: Behaviour and Interplay in Steep Turbulent Streams

Ivan Pascal

In mountain regions, steep streams play an important role in water and sediment connectivity. In these highly dynamic systems, water flow features, sediment fluxes and stream morphologies are tightly interlinked over a broad range of temporal and spatial scales. Understanding their interactions is key to forecast, prevent and mitigate the impacts of stream-related hazards on human population and facilities as well as to manage the resources connected to the rivers. Sediment clasts at rest on the bed surface may be eroded, transported and deposited by the water flow. Grains that move in rolling and saltation regimes close to the bed compose the 'bedload', which is considered the main mode of sediment transport when studying morphodynamics of mountain rivers. The fate of these particles is of primary interest for river scientists and engineers as bedload predominantly drives the river morphology evolution and responds to it in a complex manner. Therefore, measuring and predicting bedload transport rates are fundamental concerns for many works of research and applications. Bedload transport rates may exhibit large non-Gaussian fluctuations, even under steady-state conditions, especially when bedload transport intensity is weak. These fluctuations may originate from different phenomena and affect the accuracy of bedload transport rate estimates. Particles transported by the water flow can destabilise other particles at rest on the bed surface. Under weak bedload transport conditions, this positive feedback can lead to large bedload pulses. When grains settle, they often organise in small clusters or larger bedforms. Bedform migration has been identified as one of the potential causes of bedload transport fluctuations. Trains of upstream-migrating antidunes may develop in steep coarse-bedded streams when the water discharge is sufficiently high to ensure fully supercritical flow conditions. In this dissertation, I present an experimental investigation on the behaviour of bedload transport fluctuations and antidunes in steep supercritical flows. The experiments were conducted in narrow flumes under well-controlled conditions with high-resolution image-based bedload transport and bed topography monitoring. The dependence of mean bedload transport rate uncertainty on the averaging time interval was analysed. Furthermore, I performed experimental runs with increasing transport intensity and a nearly constant mean bed slope angle to study how antidune sequences composed by bedforms of variable shape typically responded to the mean bedload flux imposed under steady-state conditions. Spectral analysis of the bed elevation perturbation in time and space allowed us to quantify the variability ranges of antidune shape and celerity. Scaling the spectra with opportune reference scales made it possible to identify a consistent dimensionless relationship between antidune geometry and migration celerity, which includes the dependencies on flow and sediment supply forcings. I studied how the bedload transport rate fluctuates in presence of migrating bedforms. The effects on the bedload transport rate periodicity due to non-uniformities in the antidune migration period and to deviations from the steady-state transport conditions were also investigated. This work is a further step on the way towards a complete understanding of the interplay between bedload transport fluctuations and bedforms in mountain streams.