I-Sheng Yang

Causal machine-learning is about predicting the net-effect (true-lift) of treatments. Given the data of a treatment group and a control group, it is similar to a standard supervised-learning problem. Unfortunately, there is no similarly well-defined loss function due to the lack of point-wise true values in the data. Many advances in modern machine-learning are not directly applicable due to the absence of such loss function. We propose a novel method to define a loss function in this context, which is equal to mean-square-error (MSE) in a standard regression problem. Our loss function is universally applicable, thus providing a general standard to evaluate the quality of any model/strategy that predicts the true-lift. We demonstrate that despite its novel definition, one can still perform gradient descent directly on this loss function to find the best fit. This leads to a new way to train any parameter-based model, such as deep neural networks, to solve causal machine-learning problems without going through the meta-learner strategy.

"The chance to win given a certain move" is an easily obtainable quantity from data and often quoted in gaming statistics. It is also the fundamental quantity that reinforcement learning AI bases on. Unfortunately, this conditional probability can be misleading. Unless all players are equally skilled, this number does not tell us the intrinsic value of such move. That is because conditioning on one good move also inevitably selects a subset of better players. They tend to make other good moves, which also contribute to the extra winning chance. We present a simple toy model to quantify this "skill bias" effect, and then propose a general method to remove it. Our method is modular, generalizable, and also only requires easily obtainable quantities from data. In particular, it gets the same answer independent of whether the data comes from a group of good or bad players. This may help us to eventually break free from the conventional wisdom of "learning from the experts" and avoid the Group Thinking pitfall.

We derive the time scale for two initially pure subsystems to become entangled with each other through an arbitrary Hamiltonian that couples them. The entanglement timescale is inversely proportional to the “correlated uncertainty” between the two subsystems, a quantity which we will define and analyze in this paper. Our result is still applicable when one of the subsystems started in an arbitrarily mixed state, thus it generalizes the well-known “decoherence time scale” while coupled to a thermal state.

We derive the timescale for two initially pure subsystems to become entangled with each other through an arbitrary Hamiltonian that couples them. The entanglement timescale is inversely proportional to the "correlated uncertainty" between the two subsystems, a quantity which we will define and analyze in this paper. Our result is still applicable when one of the subsystems started in an arbitrarily mixed state, thus it generalizes the well-known "decoherence timescale" while coupled to a thermal state.

Previous work on the AdS instability problem within the two-time framework (TTF) has found an "oscillating singularity" whose presence depends on the gauge choice. We give a physical interpretation of this singularity as a diverging redshift between the boundary and the center of AdS. This signals a genuine breakdown of the linearized gravity. One can also identify the diverging redshift through a back-reaction calculation purely in the boundary gauge, where the TTF result stays regular.

Previous work on the anti–de Sitter (AdS) instability problem within the two-time framework (TTF) has found an “oscillating singularity” whose presence depends on the gauge choice. We give a physical interpretation of this singularity as a diverging redshift between the boundary and the center of AdS. This signals a genuine breakdown of the linearized gravity. One can also identify the diverging redshift through a backreaction calculation purely in the boundary gauge, where the TTF result stays regular.

We show that when a supernova explodes, a nearby pulsar signal goes through a very specific change. The observed period first changes smoothly, then is followed by a sudden change in the time derivative. A stable millisecond pulsar can allow us to measure such an effect. This may improve our measurement of the total energy released in neutrinos and also the orientation of the supernova-pulsar system.