Ali Shakiba

The development of perovskites and perovskite‐inspired materials (PIMs) is driven by the need for efficient, non‐toxic and stable solar energy conversion technologies. While halide perovskites exhibit outstanding optoelectronic properties, their practical deployment remains hindered by toxicity concerns and long‐term instability. Conventional experimental and computational approaches, though effective, are often limited by high costs and low throughput, prompting the need for data‐driven strategies. In this review, we provide a comprehensive analysis of machine learning (ML)‐driven approaches for predicting key properties such as bandgap, stability, and lattice constants in perovskite and PIMs systems. We outline a complete ML workflow, from target identification and data collection to feature engineering and model selection across supervised, unsupervised, and reinforcement learning frameworks. Special attention is given to the transferability of ML strategies developed for halide perovskites to the more chemical diverse PIMs landscape. By highlighting recent progress and current limitations, we provide a critical roadmap for integrating ML into the rational design and discovery of next‐generation non‐toxic, stable solar materials. These insights are expected to accelerate the discovery‐to‐deployment cycle for low‐toxicity, high‐efficiency solar absorbers and catalyze innovation across the broader field of data‐driven energy materials.

Photovoltaic energy is currently the most affordable form of electricity due to ongoing improvements in the efficiency and reliability of solar modules. Recently, it was suggested that 75 terawatts of installed photovoltaic capacity will be necessary by 2050 to reach the global decarbonisation targets. This demand increases the need for photovoltaic modules to become even cheaper and more reliable. The capability to predict the long‐term field performance of photovoltaic modules is vital to improving their reliability. An accurate performance prediction will also improve the bankability of utility‐scale photovoltaic plants, as trustworthy long‐term predictions reduce uncertainties in financing such projects. This study presents a proof‐of‐concept approach for predicting performance trends using a generative, deep learning model. This model was trained on time series data gathered from multiple 1512‐h accelerated damp heat tests of heterojunction PV modules. Despite using only the first 30% of the time series data, this generative learning framework accurately predicts multiple current–voltage and electroluminescence trends throughout the test duration. These predictions also highlight a key advantage over traditional approaches, which require the manual fitting of single‐parameter trends. The model is also shown to automatically correlate the impact of temperature on the degradation kinetics of photovoltaic modules, demonstrating its ability to learn and model the impact of environmental conditions on the observed degradation. While the model is trained on laboratory‐based measurements, the approach can be modified to use data collected from fielded photovoltaic modules. With further development, this approach could support improved module design and maintenance strategies, enhancing the reliability of fielded PV systems. This study is a critical step towards developing machine learning‐based tools for data‐driven reliability improvements of photovoltaic modules.

In this work, we consider a combinatorial optimization problem with direct applications in blockchain mining, namely finding the most lucrative blocks for Bitcoin miners, and propose optimal algorithmic solutions. Our experiments show that our algorithms increase the miners’ revenues by more than a million dollars per month. Modern blockchains reward their miners in two ways: (i) a base reward for each block that is mined, and (ii) the transaction fees of those transactions that are included in the mined block. The base reward is fixed by the respective blockchain’s protocol and is not under the miner’s control. Hence, for a miner who wishes to maximize earnings, the fundamental problem is to form a valid block with maximal total transaction fees and then try to mine it. Moreover, in many protocols, including Bitcoin itself, the base reward halves at predetermined intervals, hence increasing the importance of maximizing transaction fees and mining an optimal block. This problem is further complicated by the fact that transactions can be prerequisites of each other or have conflicts (in case of double-spending). In this work, we consider the problem of forming an optimal block, i.e. a valid block with maximal total transaction fees, given a set of unmined transactions. On the theoretical side, we first formally model our problem as an extension of Knapsack and then show that, unlike classical Knapsack, our problem is strongly NP-hard. We also show a hardness-of-approximation result. As such, there is no hope in solving it efficiently for general instances. However, we observe that its real-world instances are quite sparse, i.e. the transactions have very few dependencies and conflicts. Using this fact, and exploiting three well-known graph sparsity parameters, namely treedepth, treewidth and pathwidth, we present exact linear-time parameterized algorithms that are applicable to the real-world instances and obtain optimal results. On the practical side, we provide an extensive experimental evaluation demonstrating that our approach vastly outperforms the current Bitcoin miners in practice, obtaining a significant per-block average increase of 11.34 percent in transaction fee revenues which amounts to almost one million dollars per month.

Given a positive integer k and a graph G = (V, E), a function f from V to the power set of I_k is called a k-rainbow function if for each vertex v ∈ V, f(v)=∅ implies ∪_{u ∈ N(v)}f(u)=I_k where N(v) is the set of all neighbors of vertex v and I_k = {1, …, k}. Finding a k-rainbow function of minimum weight of ∑_v ∈ V|f(v)|, which is called the k-rainbow domination problem, is known to be NP-complete for arbitrary graphs and values of k. In this paper, we propose a dynamic programming algorithm to solve the k-rainbow domination problem for graphs with bounded tree-width tw in 𝒪((2^k + 1+1)^twn) time, where G has n vertices. Moreover, we also show that the same approach is applicable to solve the weighted k-rainbow domination problem with the same complexity. Therefore, both problems of k-rainbow and weighted k-rainbow domination belong to the class FPT, or fixed parameter tractable, with respect to tree-width. In addition to formally showing the correctness of our algorithms, we also implemented these algorithms to illustrate some examples.

Following the Bitcoin model, many modern blockchains reward their miners in two ways: (i) a base reward for each block that is mined, and (ii) the transaction fees of those transactions that are included in the mined block. The base reward is fixed by the respective blockchain's protocol and is not under the miner's control. Hence, for a miner who wishes to maximize earnings, the fundamental problem is to form a valid block with maximal total transaction fees and then try to mine it. Moreover, in many protocols, including Bitcoin itself, the base reward halves at predetermined intervals, hence increasing the importance of maximizing transaction fees and mining an optimal block. This problem is further complicated by the fact that transactions can be prerequisites of each other or have conflicts (in case of double-spending). In this work, we consider the problem of forming an optimal block, i.e. a valid block with maximal total transaction fees, given a set of unmined transactions. On the theoretical side, we first formally model our problem as an extension of Knapsack and then show that, unlike classical Knapsack, our problem is strongly NP-hard. We also show a hardness-of-approximation result. As such, there is no hope in solving it efficiently for general instances. However, we observe that its real-world instances are quite sparse, i.e. the transactions have very few dependencies and conflicts. Using this fact, and exploiting two well-known graph sparsity parameters, namely treewidth and pathwidth, we present exact linear-time parameterized algorithms that are applicable to the real-world instances and obtain optimal results. We also provide an experimental evaluation demonstrating that our approach vastly outperforms the current Bitcoin miners in practice, obtaining a significant per-block average increase of 13.4% in transaction fee revenues.