We demonstrate experimentally an efficient control of light intensity distribution inside a random scattering system. The adaptive wavefront shaping technique is applied to a silicon waveguide containing scattering nanostructures, and the on-chip coupling scheme enables access to all input spatial modes. By selectively coupling the incident light to open or closed channels of the disordered system, we not only vary the total energy stored inside the system by 7.4 times, but also change the energy density distribution from an exponential decay to a linear decay and to a profile peaked near the center. This work provides an on-chip platform for controlling light-matter interactions in turbid media.PACS numbers: 42.25. Bs, 42.25.Dd, It has long been known that in disordered media there are many fascinating counter-intuitive effects resulting from interferences of multiply scattered waves [1,2]. One of them is the creation of transmission eigenchannels which can be broadly classified as open and closed [3,4]. Existence of high-transmission (open) channels allows an optimally prepared coherent input beam transmitting through a lossless diffusive medium with order unity efficiency. Opposite to that, waves injected to lowtransmission (closed) channels can barely penetrate the medium and are mostly reflected instead. In general, the penetration depth and energy density distribution of multiply scattered waves inside a disordered medium are determined by the spatial profiles of the transmission eigenchannels that are excited by the incident light. The distinct spatial profiles of open and closed channels suggest that selective coupling of incident light to these channels enables an effective control of total transmission and energy distribution inside the random medium [5,6]. Since the energy density determines the light-matter interactions inside a scattering system, manipulating its spatial distribution opens the door to tailoring optical excitations as well as linear and nonlinear optical processes such as absorption, emission, amplification, and frequency mixing inside turbid media. The potential applications range from photovoltaics [7,8] In recent years there have been numerous theoretical and experimental studies on transmission eigenchannels [5,[13][14][15][16][17]. While they can be deduced from the measured transmission matrix [18][19][20][21], it is difficult to directly probe their spatial profiles inside three-dimensional (3D) random media. So far, the open and closed channels are observed only with acoustic wave inside a two-dimensional (2D) disordered waveguide [22], but controlling the energy density distribution has not been realized due to lack of an efficient wavefront modulator for acoustic wave or microwave. The advantage for optical wave is the availability of spatial light modulator (SLM) with many degrees of freedom, however, the commonly used samples in the optics experiment have an open slab geometry, making it impossible to control all input modes due to finite numerical aperture of the imaging...

The universal bimodal distribution of transmission eigenvalues in lossless diffusive systems underpins such celebrated phenomena as universal conductance fluctuations, quantum shot noise in condensed matter physics and enhanced transmission in optics and acoustics. Here, we show that in the presence of absorption, density of the transmission eigenvalues depends on the confinement geometry of scattering media. Furthermore, in an asymmetric waveguide, densities of the reflection and absorption eigenvalues also depend of the side from which the waves are incident. With increasing absorpotion, the density of absorption eigenvalues transforms from single-peak to double-peak function. Our findings open a new avenue for coherent control of wave transmission, reflection and absorption in random media.PACS numbers: 42.25. Dd, 42.25.Bs, Mesoscopic electronic transport through a disordered conductor can be described by a N × N transmission matrixt which relates the amplitudes of N incoming and outgoing transverse modes [1]. Dimensionless conductance is g = Tr t †t = n τ n , where τ n are the eigenvalues of the matrixt †t [2] and ... denotes ensemble average. Therefore, electron transport in a metallic wire can be viewed as parallel transmission over N orthogonal eigenchannels with individual transmissions of τ n . Due to the mesoscopic correlations [3,4], density of the transmission eigenvalues D(τ ) has a bimodal functional form [5][6][7][8][9][10][11] with peaks at τ → 0 and τ → 1. The latter implies the existence of nearly perfect transmission eigenchannels [5,12, 13] that lead to e.g. universal conductance fluctuations [14, 15] and quantum shot noise [16,17]. In Ref.[18], bimodal distribution was proven to be applicable to an arbitrary geometry of the conductor as long as the transport remains diffusive and free of dissipation.The bimodal distribution obtained in the context of mesoscopic physics is also applicable to transport of classical waves in scattering media [19]. In optics, rapid development of spatial light modulators has enabled an experimental access to the transmission eigenchannels [20] that led to experimental demonstration of enhanced transmission [21,22] with applications in focusing and imaging through turbid media [21,[23][24][25][26][27][28][29][30]. Absorption, common in optics, breaks energy conservation and makes the density of transmission eigenvalues [31] as well as reflection [32][33][34] eigenvalues to depend on its strength. However, the questions of whether the geometry of the system could affect the eigenvalue density in dissipative systems and if so, how it would affect it, have not been addressed.In this work we demonstrate that, unlike passive systems, the density of the transmission eigenvalues in absorbing disordered waveguides is geometry dependent, that is beyond predictions of the existing theory [31]. This opens possibility of tuning the functional form of the eigenvalue density by choosing the shape of the boundary. Furthermore, we show that dissipation makes a profound impact o...

<p>Context recognition for lifelong learning (L2) agents is an open-ended problem whereby aggregate features in an environment are utilized to signal the active context in which the agent is operating. The ability to recognize context is necessary in L2 agents to engage modulatory signals to account for significant changes in the input state space associated with a given context or task, such as altering learning dynamics or shifting attention to more relevant features. Context recognition is itself an L2 problem due to the ever-increasing number of distinct contexts that an agent might encounter, requiring incrementally learning novel contexts while prescribing them to supervised task labels when available. This paper demonstrates an algorithm based on near clustering of deep-extracted features with adaptive resonance theory methods that satisfies these requirements on the behalf of an embodied L2 agent in a computer vision environment. The strength of this algorithm lies in its flexibility, being capable of online incremental learning in supervised, realistic semi-supervised, and unsupervised scenarios while demonstrating continual learning in its own right.</p>

<p>Understanding the performance and validity of clustering algorithms is both challenging and crucial, particularly when clustering must be done online. Until recently, most validation methods have relied on batch calculation and have required considerable human expertise in their interpretation. Improving real-time performance and interpretability of cluster validation, therefore, continues to be an important theme in improving unsupervised learning. Building upon previous work on incremental cluster validity indices (iCVIs), this paper introduces the Meta-iCVI as a tool for explainable and concise labeling of partition quality in online clustering. Some iCVIs are better at detecting under-partition; others at over-partition. Combining them was hypothesized to improve cluster validation analysis. Experiments were conducted on generalized synthetic and real-world data sets to demonstrate the efficacy and application of this method.</p> <p>Results of 100% accuracy were achieved in labeling partition quality on real-world data sets including MNIST and FLIR ADAS, demonstrating that the Meta-iCVI is a powerful and efficient tool for classifying partition quality in a variety of conditions. Its introduction should empower new and more efficient streaming clustering techniques. Additionally, we believe this to be the first implementation of an ensemble iCVI metric and the first time iCVI validation performance has been evaluated on randomized sample presentation.</p>

No abstract