Talks

Bayesian causal structure learning aims to learn a posterior distribution over directed acyclic graphs (DAGs), and the mechanisms that define the relationship between parent and child variables. By taking a Bayesian approach, it is possible to reason about the uncertainty of the causal model. The notion of modelling the uncertainty over models is particularly crucial for causal structure learning since the model could be unidentifiable when given only a finite amount of observational data. In this paper, we introduce a novel method to jointly learn the structure and mechanisms of the causal model using Variational Bayes, which we call Variational Bayes-DAG-GFlowNet (VBG). We extend the method of Bayesian causal structure learning using GFlowNets to learn not only the posterior distribution over the structure, but also the parameters of a linear-Gaussian model. Our results on simulated data suggest that VBG is competitive against several baselines in modelling the posterior over DAGs and mechanisms, while offering several advantages over existing methods, including the guarantee to sample acyclic graphs, and the flexibility to generalize to non-linear causal mechanisms.

---

Zoom link 

The Bay of Bengal (bay) is a semi-enclosed tropical basin driven by seasonally reversing monsoon winds and a huge quantity of freshwater from rainfall and river runoff. The bay plays a fundamental role in controlling weather systems that make up the Asian summer monsoon system, including monsoon depressions and tropical cyclones. We have used a high resolution (~1 km) regional ocean model of the Bay of Bengal to explore the sub-mesoscale variability in the bay. Model simulations show that the East India Coastal Current (EICC) is extremely rich in submesoscale features compared to the open ocean and exhibit significant seasonal variations. Submesoscale activity over the EICC region is weakest during spring (March-May), slightly stronger during summer monsoon (June-September) and strongest during winter monsoon (November-January). Weak winds during spring and a huge fresh-water gain during summer monsoon tend to weaken submesoscale activity. Investigation of conversion rates of APE to KE...

Surface currents modify wind driven Ekman pumping in the ocean both by modifying the stress itself and by modifying the relationship between the stress and the Ekman transport. The former effect results in a strong mesoscale structure in the wind stress curl, such as is evident from scatterometer data. This mesoscale forcing is anti-correlated with surface vorticity and thus produces a strong damping effect on ocean eddies and currents. Recent work, however, suggests that this damping effect is over-represented in common parameterizations of the air-sea wind stress. The latter effect is referred to as nonlinear Ekman dynamics. These dynamics take the stress as given and add advective terms to the linear balance. Specifically, cross terms involving the Ekman and non-Ekman components of the flow are added to the linear Ekman balance. This is known to produce small scale (e.g., submesoscale) structures in the pumping velocity.

Here, we first review both the ocean surface velocity depen...

Surface currents modify wind driven Ekman pumping in the ocean both by modifying the stress itself and by modifying the relationship between the stress and the Ekman transport. The former effect results in a strong mesoscale structure in the wind stress curl, such as is evident from scatterometer data. This mesoscale forcing is anti-correlated with surface vorticity and thus produces a strong damping effect on ocean eddies and currents. Recent work, however, suggests that this damping effect is over-represented in common parameterizations of the air-sea wind stress. The latter effect is referred to as nonlinear Ekman dynamics. These dynamics take the stress as given and add advective terms to the linear balance. Specifically, cross terms involving the Ekman and non-Ekman components of the flow are added to the linear Ekman balance. This is known to produce small scale (e.g., submesoscale) structures in the pumping velocity.

Here, we first review both the ocean surface velocity depen...

Critical behaviour associated with the occurrence of tipping points is one of the key aspects of the dynamics of the Earth system. Criticality occurs in many different forms and with different characteristic spatial and temporal scales. Past critical transitions have sometimes been associated with mass extinction events, and the snowball/warm Earth dichotomy has played a major role in the emergence of multicellular life. Currently, anthropogenic forcing to the climate system seem to be bringing some multi stable subcomponents of the climate closer to a bifurcation point, as in the case of the Atlantic meridional overturning circulation and the Greenland ice sheet. We will present here a general mathematical framework to look into critical behaviour in the Earth system. We will connect the analysis of the response of the system to perturbation in connection to its global stability properties and discuss ideas for creating robust early warning signals.

Recent efforts in building data-driven surrogates for weather forecasting applications have received a lot of attention and garnered noticeable success. These autoregressive data-driven models yield significantly competitive short-term forecasting performance (often outperforming traditional numerical weather models) at a fraction of the computational cost of numerical models. However, these data-driven models do not remain stable when time-integrated for a long time. Such a long time-integration would provide (1) a method to seamlessly scale a weather model to a climate model and (2) gathering insights into the statistics of that climate system, e.g., the extreme events, owing to the cheap cost of generating multiple ensembles. While many studies have reported this instability, especially for data-driven models of turbulent flow, a causal mechanism for this instability is not clear. Most efforts to obtain stability are ad-hoc and empirical. In this work, we use a canonical quasi-geost...

I will argue that there are quantum states of the field theories of general relativity and electromagnetism that we typically ignore, but have interesting phenomenological effects. These states amount to relaxing the constraint equations known as the Hamiltonian and momentum constraints in GR and Gauss’ law in EM. Turning off the Hamiltonian constraint sources non-dynamical parts of the metric which mimic a pressure-less dust, and thus these effects may be the explanation as to why we have inferred the existence of dark matter, both locally and cosmologically. Turning off the momentum constraints add additional velocity-dependent source terms to this effective dust, but these effects are not conserved and redshift quickly outside the horizon. Turning off the Gauss’ law constraint mimics a charge density that does not respond to electric forces, but follows geodesics, thus adding a charged component to the dust. The effects in electromagnetism may have interesting impacts on BBN, the baryon-photon fluid during and after recombination, galactic dynamics, and cosmic rays. If this new structure in the gravitational and electric fields explain dark matter, it forbids an early period of inflation and therefore requires a different explanation for density perturbations.