Accurate Multi-Objective Design in a Space of Millions of Transition Metal Complexes with **Neural**-Network-Driven Efficient Global Optimization

**neural**

**network**, ANN, or Gaussian process, GP) models for this task are limited by training data availability and predictive uncertainty quantification (UQ). Expand abstract.

**neural**network, ANN, or Gaussian process, GP) models for this task are limited by training data availability and predictive uncertainty quantification (UQ). We overcome such limitations by using efficient global optimization (EGO) with the multi-dimensional expected improvement (EI) criterion. EGO balances exploitation of a trained model with acquisition of new DFT data at the Pareto front, the region of chemical space that contains the optimal trade-off between multiple design criteria. We demonstrate this approach for the simultaneous optimization of redox potential and solubility in candidate M(II)/M(III) redox couples for redox flow batteries from a space of 2.8M transition metal complexes designed for stability in practical RFB applications. We employ latent-distance-based UQ with a multi-task ANN to enable model generalization that surpasses that of a GP. With this approach, ANN prediction and EI scoring of the full 2.8M complex space is achieved in minutes. Starting from ca. 100 representative points, EGO improves both properties by 3-4 standard deviations in only five generations. Analysis of lookahead errors confirms rapid ANN model improvement during the EGO process, achieving suitable accuracy for predictive design in the space of transition metal complexes. The ANN-driven EI approach achieves at least 500-fold acceleration over random search, identifying a Pareto-optimal design in around five weeks instead of fifty years.

5/10 relevant

chemRxiv

Integrating Deep **Neural** **Networks** and Symbolic Inference for Organic Reactivity Prediction

**neural**

**networks**by predicting products with more than 90% accuracy, demonstrates intuitive chemical reasoning through a learned attention mechanism, and provides generalizability across various reaction types. Expand abstract.

**neural**

**networks**with a probabilistic and symbolic inference that flexibly enforces chemical constraints and accounts for prior chemical knowledge. We first train a graph convolutional

**neural**

**network**to estimate the likelihood of changes in covalent bonds, hydrogen counts, and formal charges. These estimated likelihoods govern a probability distribution over potential products. Integer Linear Programming is then used to infer the most probable products from the probability distribution subject to heuristic rules such as the octet rule and chemical constraints that reflect a user's prior knowledge. Our approach outperforms previous graph-based

**neural**

**networks**by predicting products with more than 90% accuracy, demonstrates intuitive chemical reasoning through a learned attention mechanism, and provides generalizability across various reaction types. Furthermore, we demonstrate the potential for even higher model accuracy when complemented by expert chemists contributing to the system, boosting both machine and expert performance. The results show the advantages of empowering deep learning models with chemical intuition and knowledge to expedite the drug discovery process.

10/10 relevant

chemRxiv

Discovery of Novel Chemical Reactions by Deep Generative Recurrent **Neural** **Network**

5/10 relevant

chemRxiv

DenseCPD: Improving the Accuracy of **Neural**-Network-Based Computational Protein Sequence Design with DenseNet

**neural**

**network**named DenseCPD to perform computational protein design given a backbone structure of a protein. Expand abstract.

**neural**

**network**named DenseCPD to perform computational protein design given a backbone structure of a protein. The uploaded files include a main manuscript and a supporting file.

7/10 relevant

chemRxiv

An Autoencoder and Artificial **Neural** **Network**-based Method to Estimate Parity Status of Wild Mosquitoes from Near-infrared Spectra

**neural**

**network**(ANN) models on NIR spectra to estimate the parity status of wild mosquitoes. Expand abstract.

**neural**

**network**(ANN) models on NIR spectra to estimate the parity status of wild mosquitoes. We use four different datasets: An. arabiensis collected from Minepa, Tanzania (Minepa-ARA); An. gambiae s.s collected from Muleba, Tanzania (Muleba-GA); An. gambiae s.s collected from Burkina Faso (Burkina-GA); and An.gambiae s.s from Muleba and Burkina Faso combined (Muleba-Burkina-GA). We train ANN models on datasets with spectra preprocessed according to previous protocols. We then use autoencoders to reduce the spectra feature dimensions from 1851 to 10 and re-train the ANN models. Before the autoencoder was applied, ANN models estimated parity status of mosquitoes in Minepa-ARA, Muleba-GA, Burkina-GA and Muleba-Burkina-GA with out-of-sample accuracies of 81.9 {+/-} 2.8 (N=274), 68.7 {+/-} 4.8 (N=43), 80.3 {+/-} 2.0 (N=48), and 75.7 {+/-} 2.5 (N=91), respectively. With the autoencoder, ANN models tested on out-of-sample data achieved 97.1 {+/-} 2.2% (N=274), 89.8 {+/-} 1.7% (N=43), 93.3 {+/-} 1.2% (N=48), and 92.7 {+/-} 1.8% (N=91) accuracies for Minepa-ARA, Muleba-GA, Burkina-GA, and Muleba-Burkina-GA, respectively. These results show that a combination of an autoencoder and an ANN trained on NIR spectra to estimate the parity status of wild mosquitoes yields models that can be used as an alternative tool to estimate parity status of wild mosquitoes, especially since NIRS is a high-throughput, reagent-free, and simple-to-use technique compared to ovary dissections.

7/10 relevant

bioRxiv

Epigenetic engineering of yeast reveals dynamic molecular adaptation to methylation stress and genetic modulators of specific DNMT3 family members

**neural**

**networks**trained on genome-wide CpG-methylation data learned distinct sequence preferences of DNMT3 family members. Expand abstract.

**network**-level adaptation of cells to DNMT3B1-induced DNA methylation stress and showed that coordinately modulating the availability of S-adenosyl methionine (SAM), the essential metabolite for DNMT-catalyzed methylation, is an evolutionarily conserved epigenetic stress response, also implicated in several human diseases. Convolutional

**neural**

**networks**trained on genome-wide CpG-methylation data learned distinct sequence preferences of DNMT3 family members. A simulated annealing interpretation method resolved these preferences into individual flanking nucleotides and periodic poly(A) tracts that rotationally position highly methylated cytosines relative to phased nucleosomes. Furthermore, the nucleosome repeat length defined the spatial unit of methylation spreading. Gene methylation patterns were similar to those in mammals, and hypo- and hypermethylation were predictive of increased and decreased transcription relative to control, respectively, in the absence of mammalian readers of DNA methylation. Introducing controlled epigenetic perturbations in yeast thus enabled characterization of fundamental genomic features directing specific DNMT3 proteins.

4/10 relevant

bioRxiv

Estimation for Compositional Data using Measurements from Nonlinear
Systems using Artificial **Neural** **Networks**

**neural**

**networks**(ANNs) can compete with the optimal bounds for linear systems, where convex optimization theory applies, and demonstrate promising results for nonlinear system inversions. Expand abstract.

**neural**

**networks**(ANNs) can compete with the optimal bounds for linear systems, where convex optimization theory applies, and demonstrate promising results for nonlinear system inversions. We performed extensive experiments by designing numerous different types of nonlinear systems.

10/10 relevant

arXiv

Low-Complexity LSTM Training and Inference with FloatSD8 Weight Representation

**neural**

**networks**(RNNs), specifically long short-term memory (LSTM). Expand abstract.

**neural**

**networks**(CNNs) training and inference. In this paper, we applied FloatSD to recurrent

**neural**

**networks**(RNNs), specifically long short-term memory (LSTM). In addition to FloatSD weight representation, we quantized the gradients and activations in model training to 8 bits. Moreover, the arithmetic precision for accumulations and the master copy of weights were reduced from 32 bits to 16 bits. We demonstrated that the proposed training scheme can successfully train several LSTM models from scratch, while fully preserving model accuracy. Finally, to verify the proposed method's advantage in implementation, we designed an LSTM neuron circuit and showed that it achieved significantly reduced die area and power consumption.

5/10 relevant

arXiv

Ada-LISTA: Learned Solvers Adaptive to Varying Models

**Neural**

**networks**that are based on unfolding of an iterative solver, such as LISTA (learned iterative soft threshold algorithm), are widely used due to their accelerated performance. Expand abstract.

**Neural**

**networks**that are based on unfolding of an iterative solver, such as LISTA (learned iterative soft threshold algorithm), are widely used due to their accelerated performance. Nevertheless, as opposed to non-learned solvers, these

**networks**are trained on a certain dictionary, and therefore they are inapplicable for varying model scenarios. This work introduces an adaptive learned solver, termed Ada-LISTA, which receives pairs of signals and their corresponding dictionaries as inputs, and learns a universal architecture to serve them all. We prove that this scheme is guaranteed to solve sparse coding in linear rate for varying models, including dictionary perturbations and permutations. We also provide an extensive numerical study demonstrating its practical adaptation capabilities. Finally, we deploy Ada-LISTA to natural image inpainting, where the patch-masks vary spatially, thus requiring such an adaptation.

4/10 relevant

arXiv

DCT-Conv: Coding filters in convolutional **networks** with Discrete Cosine
Transform

**neural**

**networks**are based on a huge number of trained weights. Consequently, they are often data-greedy, sensitive to overtraining, and learn slowly. We follow the line of research in which filters of convolutional

**neural**layers are determined on the basis of a smaller number of trained parameters. In this paper, the trained parameters define a frequency spectrum which is transformed into convolutional filters with Inverse Discrete Cosine Transform (IDCT, the same is applied in decompression from JPEG). We analyze how switching off selected components of the spectra, thereby reducing the number of trained weights of the network, affects its performance. Our experiments show that coding the filters with trained DCT parameters leads to improvement over traditional convolution. Also, the performance of the

**networks**modified this way decreases very slowly with the increasing extent of switching off these parameters. In some experiments, a good performance is observed when even 99.9% of these parameters are switched off.

6/10 relevant

arXiv