We introduce the Brain Language Model (BrainLM), a foundation model for brain activity dynamics trained on 6,700 hours of fMRI recordings. Utilizing self-supervised masked-prediction training, BrainLM demonstrates proficiency in both fine-tuning and zero-shot inference tasks. Fine-tuning allows for the accurate prediction of clinical variables like age, anxiety, and PTSD as well as forecasting of future brain states. Critically, the model generalizes well to entirely new external cohorts not seen during training. In zero-shot inference mode, BrainLM can identify intrinsic functional networks directly from raw fMRI data without any network-based supervision during training. The model also generates interpretable latent representations that reveal relationships between brain activity patterns and cognitive states. Overall, BrainLM offers a versatile and interpretable framework for elucidating the complex spatiotemporal dynamics of human brain activity. It serves as a powerful "lens" through which massive repositories of fMRI data can be analyzed in new ways, enabling more effective interpretation and utilization at scale. The work demonstrates the potential of foundation models to advance computational neuroscience research.

Compound channel models offer a simple and straightforward way of analyzing the stability of decoder design under model variations. With this work we provide a coding theorem for a large class of practically relevant compound channel models. We give explicit formulas for the cases of the Gaussian classical-quantum compound channels with unknown noise, unknown phase and unknown attenuation. We show analytically how the classical compound channel capacity formula motivates nontrivial choices of the displacement parameter of the Kennedy receiver. Our work demonstrates the value of the compound channel model as a method for the design of receivers in quantum communication.

We evaluate the performance of coherent frequency-shift keying (CFSK) [1], [2] alphabets for communication on quantum channels. We show that, contrarily to what previously thought, the square-root-measurement (SRM) is sub-optimal for discriminating CFSK states. Furthermore, we compute the maximum information transmission rate of the CFSK alphabet, observing that it employs at least as many frequency modes as the signal states, and compare it with standard phase-shift-keying. Finally, we introduce a discretized CFSK alphabet with higher mode-efficiency, exhibiting comparable error-probability performance with respect to CFSK and better rate performance. Our results suggest the existence of a tradeoff between the CFSK reduced error-probability and its mode efficiency.

We study the problem of calibrating a quantum receiver for optical coherent states when transmitted on a quantum optical channel with variable transmissivity, a common model for long-distance optical-fiber and free/deep-space optical communication [1]–[7]. We optimize the error probability of legacy adaptive receivers, such as Kennedy’s and Dolinar’s [8], [9], on average with respect to the channel transmissivity distribution. We then compare our results with the ultimate error probability attainable by a general quantum device, computing the Helstrom bound for mixtures of coherent-state hypotheses, for the first time to our knowledge, and with homodyne measurements. With these tools, we first analyze the simplest case of two different transmissivity values; we find that the strategies adopted by adaptive receivers exhibit strikingly new features as the difference between the two transmissivities increases. Finally, we employ a recently introduced library of shallow reinforcement learning methods [10], demonstrating that an intelligent agent can learn the optimal receiver setup from scratch by training on repeated communication episodes on the channel with variable transmissivity and receiving rewards if the coherent-state message is correctly identified.

We study the problem of transmitting classical information on a quantum channel in the absence of a shared phase reference. This problem is relevant for long-distance communications in free space and optical fiber, where phase noise is typically considered as a limiting factor. Previous analyses considered phase noise that acts independently on each communication mode, thus completely decohering successive signals and making it impossible to establish a phase reference. In the present work we analyze instead the realistic case in which the phase reference is lost only after m uses of the transmission line, due to a finite decoherence time. In this setting, focusing on the simplest case m = 2, we analyze two communication strategies using coherent states of the electromagnetic field and show that it is not beneficial to employ the total energy to establish a reference frame but rather to spread out the energy on all the modes and directly transmit information using their relative degrees of freedom.