Integrated Calendar

<p>Submarine groundwater discharge (SGD) is a globally important process</p><p>supplying nutrients and trace elements to the coastal environment, thus</p><p>playing a pivotal role in sustaining marine primary productivity. Along with</p><p>nutrients, groundwater also contains allochthonous microbes that are</p><p>discharged from the terrestrial subsurface into the sea. Currently, little is</p><p>known about the interactions between groundwater-borne and coastal</p><p>seawater microbial populations, and groundwater microbes' role upon</p><p>introduction to coastal seawater populations. In the current study we</p><p>investigated seawater microbial abundance, activity and diversity in a site</p><p>strongly influenced by SGD. In addition, through laboratory-controlled</p><p>bottle incubations, we mimicked different mixing scenarios between</p><p>groundwater and seawater. Our results demonstrate that the addition of</p><p>0.1 μm filtered groundwater stimulated heterotrophic activity and</p><p>increased microbial abundance compared to control coastal seawater,</p><p>whereas 0.22 μm filtration treatments induced primary productivity and</p><p>Synechococcus growth. 16S rRNA gene sequencing showed a strong</p><p>shift from a SAR11-rich community in the control samples to</p><p>Rhodobacteraceae dominance in the &lt;0.1 μm treatment, in agreement</p><p>with Rhodobacteraceae enrichment in the SGD field site. These results</p><p>suggest that microbes delivered by SGD may affect the abundance,</p><p>activity and diversity of intrinsic microbes in coastal seawater, highlighting</p><p>the cryptic interplay between groundwater and seawater microbes in</p><p>coastal environments, which has important implications for carbon</p><p>cycling.</p>

<p>Despite extensive studies at the genomic, transcriptomic, and metabolomic levels, the underlying mechanisms regulating longevity remain incompletely understood. Post-translational protein acetylation has been suggested to regulate aspects of longevity. To further explore the role of acetylation, we developed the PHARAOH computational tool, based on the 100-fold differences in longevity within the mammalian class. Analyzing acetylome and proteome data across 107 mammalian species identified multiple significant longevity-associated acetylated lysine residues in mice and humans, controlling many longevity-related pathways. Specifically, we found that longevity-associated acetylation sites help resolve the Peto Paradox: the enigma of why animals with increased body size live longer yet do not exhibit much higher cancer incidence. Our findings show a significant positive correlation between these new acetylation sites and protection against multiple types of cancer in humans. Moreover, mutating these sites reduced the anti-neoplastic functions of the acetylated proteins. These findings provide new insights into the pivotal role of protein acetylation in mammalian longevity, suggesting potential interventions to extend human healthspan.</p>

The role of the algorithm in generalization remains one of the least understood aspects of modern machine learning. Classical theories, such as VC-theory and PAC learning, posits that the sample size needed to fit a model depends only on the class to be learnt but not on the fitting algorithm itself. Yet, in practice, the algorithm plays a crucial role in avoiding overfitting. A well-studied framework to explore this possibility is Stochastic Convex Optimization, where the algorithm's influence on generalization is well established. We will discuss two recent results that try to shed light on how algorithms affect generalization.

The first result examines the sample complexity of Gradient Descent. Arguably this is one of the simplest algorithms for this setup. We will present the first tight sample complexity bounds. These bounds demonstrate how, when applied naively, Gradient Descent performs no better than a worst-case empirical risk minimizer. However, with correct parameter tuning, the algorithm achieves optimal sample complexity rates, but in a computationally inefficient manner.

The second result investigates the interplay between memorization and learning and the potential of information-theoretic generalization bounds. Contrary to the conventional view that successful learning avoids memorization, we will see that even in simple scenarios, memorization can be essential. This finding suggests that large-scale learning might, unintuitively, require complete memorization of the dataset.

<p>Join our ABC CHATS where CEO's share their ABC's</p><p>on scientific leadership, breakthroughs and failures throughout their personal stories </p>

Nonlinear activation functions are pivotal to the success of deep neural nets, and choosing the appropriate activation function can significantly affect their performance. Most networks use fixed activation functions (e.g., ReLU, GELU, etc.), and this choice might limit their expressiveness. Furthermore, different layers may benefit from diverse activation functions. Consequently, there has been a growing interest in trainable activation functions. In this paper, we introduce DiTAC, a trainable highly-expressive activation function based on an efficient diffeomorphic transformation (called CPAB). Despite introducing only a negligible number of trainable parameters, DiTAC enhances model expressiveness and performance, often yielding substantial improvements. It also outperforms existing activation functions (regardless whether the latter are fixed or trainable) in tasks such as semantic segmentation, image generation, regression problems, and image classification. The talk is based on [Chelly et. all, ECCV '24].

Paper:

https://arxiv.org/abs/2407.07564

Speakers' short bio:

Irit Chelly is a PhD student in the Computer Science department at Ben-Gurion University, where she also earned her M.Sc., under the supervision of Dr. Oren Freifeld in the Vision, Inference, and Learning group. Her research focuses on probabilistic clustering using non-parametric Bayesian models and unsupervised learning. Her previous projects involved spatial transformations and dimensionality reduction in video analysis, and generative models. Irit won the national-level Aloni PhD scholarship from Israel's Ministry of Technology and Science as well as the BGU Hi-tech scholarship for excellent PhD students, and received annually awards and instructor rank for outstanding teaching skills in essential courses in the Computer Science department. 

Shira Ifergane is an MSc Computer Science student at BGU, working at the Vision, Inference, and Learning group under the supervision of Prof. Oren Freifeld. Shira co-authored an ECCV 2024 paper and has won the national MS scholarship for AI and Data Science research from Israel's Council for Higher Education. Her current research centers on efficient deep models for video analysis.

<p>In nature, sequence-specific biopolymers, such as peptides and nucleic acids, are essential to various biological systems and processes. These biopolymers are utilized in materials science to achieve precise property control. Typically, variations in amino acid sequences focus on functional regulation while nucleotides are used for structural control. This raises the question: How can we integrate peptide-based functionality with the spatial precision of DNA nanotechnology for innovative material design? Here, I will present examples illustrating the incredible properties of peptide self-assembly from my PhD, and the remarkable nanoarchitecture design achieved through DNA nanotechnology from my Postdoc. These two key elements establish a vision of utilizing and synergizing peptide functionality with structural control achieved by DNA nanotechnology.</p><p>Specifically, I will show how subtle changes in the molecular environment influence the morphology and behavior of peptide assemblies such as diphenylalanine crystals and enable control over their growth and disassembly processes, revealing insights into peptide-based material manipulation (Nat. Commun., 2016). Another example is that of the amorphous assemblies of tri-tyrosine peptides, where we linked the molecular arrangement to unique mechanical and optical properties of glass-like peptide structures (Nature, 2024).</p><p>Next, I will introduce the principles of DNA nanotechnology for advanced structural control. By designing DNA nano-frames capable of self-assembling into organized lattices, we created micron-scale 3D materials. We discovered that a minor modification in DNA linker length induces a crystalline phase transition, from simple cubic to face-centered cubic structures, altering lattice geometry. In addition, we established a method using acoustic waves to achieve scalable and morphologically controllable DNA assemblies at the millimetric scale (Nat. Commun., 2024). This approach highlights how DNA nanotechnology provides unparalleled spatial control, decoupling structural architecture from functional elements such as peptides and nanoparticles. Together, these projects illustrate how peptides and DNA nanotechnology can be potentially integrated to engineer novel materials and enhance our capacity to design materials with tailored properties across scales.</p>

<p>The study of dilute, strongly interacting quantum gases reveals</p><p>universal properties that transcend the specifics of individual</p><p>systems. These features arise from their short-range behavior</p><p>and are encapsulated in a key quantity called the “contact”, which</p><p>quantifies the probability of two particles being in close proximity.</p><p>In this talk, I will introduce the contact theory and its extension to</p><p>nuclear and molecular systems beyond the zero-range limit. I will</p><p>demonstrate its applicability in analyzing nuclear electron</p><p>scattering and photo absorption reactions.</p><p>Additionally, I will discuss how mean-field approximations, such as</p><p>the nuclear shell model, can effectively estimate the contact,</p><p>offering valuable insights into the underlying physics.</p>