The question is posted on Cross Validated (haven't figured out how to write formulas in reddit). Feel free to leave your comments either here or at Cross Validated.

In this work, we describe a set of rules for the design and initialization of well-conditioned neural networks, guided by the goal of naturally balancing the diagonal blocks of the Hessian at the start of training. Our design principle balances multiple sensible measures of the conditioning of neural networks. We prove that for a ReLU-based deep multilayer perceptron, a simple initialization scheme using the geometric mean of the fan-in and fan-out satisfies our scaling rule. For more sophisticated architectures, we show how our scaling principle can be used to guide design choices to produce well-conditioned neural networks, reducing guess-work.

Machine learning has recently been widely adopted to address the managerial decision making problems. However, there is a trade-off between performance and interpretability. Full complexity models (such as neural network-based models) are non-traceable black-box, whereas classic interpretable models (such as logistic regression) are usually simplified with lower accuracy. This trade-off limits the application of state-of-the-art machine learning models in management problems, which requires high prediction performance, as well as the understanding of individual attributes' contributions to the model outcome. Multiple criteria decision aiding (MCDA) is a family of interpretable approaches to depicting the rationale of human decision behavior. It is also limited by strong assumptions (e.g. preference independence). In this paper, we propose an interpretable machine learning approach, namely Neural Network-based Multiple Criteria Decision Aiding (NN-MCDA), which combines an additive MCDA model and a fully-connected multilayer perceptron (MLP) to achieve good performance while preserving a certain degree of interpretability. NN-MCDA has a linear component (in an additive form of a set of polynomial functions) to capture the detailed relationship between individual attributes and the prediction, and a nonlinear component (in a standard MLP form) to capture the high-order interactions between attributes and their complex nonlinear transformations. We demonstrate the effectiveness of NN-MCDA with extensive simulation studies and two real-world datasets. To the best of our knowledge, this research is the first to enhance the interpretability of machine learning models with MCDA techniques. The proposed framework also sheds light on how to use machine learning techniques to free MCDA from strong assumptions.

We propose a novel, conceptually simple and general framework for instance segmentation on 3D point clouds. Our method, called 3D-BoNet, follows the simple design philosophy of per-point multilayer perceptrons (MLPs). The framework directly regresses 3D bounding boxes for all instances in a point cloud, while simultaneously predicting a point-level mask for each instance. It consists of a backbone network followed by two parallel network branches for 1) bounding box regression and 2) point mask prediction. 3D-BoNet is single-stage, anchor-free and end-to-end trainable. Moreover, it is remarkably computationally efficient as, unlike existing approaches, it does not require any post-processing steps such as non-maximum suppression, feature sampling, clustering or voting. Extensive experiments show that our approach surpasses existing work on both ScanNet and S3DIS datasets while being approximately 10x more computationally efficient. Comprehensive ablation studies demonstrate the effectiveness of our design.

Neural network learning is typically slow since backpropagation needs to compute full gradients and backpropagate them across multiple layers. Despite its success of existing work in accelerating propagation through sparseness, the relevant theoretical characteristics remain unexplored and we empirically find that they suffer from the loss of information contained in unpropagated gradients. To tackle these problems, in this work, we present a unified sparse backpropagation framework and provide a detailed analysis of its theoretical characteristics. Analysis reveals that when applied to a multilayer perceptron, our framework essentially performs gradient descent using an estimated gradient similar enough to the true gradient, resulting in convergence in probability under certain conditions. Furthermore, a simple yet effective algorithm named memorized sparse backpropagation (MSBP) is proposed to remedy the problem of information loss by storing unpropagated gradients in memory for the next learning. The experiments demonstrate that the proposed MSBP is able to effectively alleviate the information loss in traditional sparse backpropagation while achieving comparable acceleration.

There is a mismatch between the standard theoretical analyses of statistical machine learning and how learning is used in practice. The foundational assumption supporting the theory is that we can represent features and models using real-valued parameters. In practice, however, we do not use real numbers at any point during training or deployment. Instead, we rely on discrete and finite quantizations of the reals, typically floating points. In this paper, we propose a framework for reasoning about learning under arbitrary quantizations. Using this formalization, we prove the convergence of quantization-aware versions of the Perceptron and Frank-Wolfe algorithms. Finally, we report the results of an extensive empirical study of the impact of quantization using a broad spectrum of datasets.

The quantity of labelled data is small due to privacy concerns and the cost of data acquisition and labelling by a medical expert. Furthermore, it is quite common that collected data are unbalanced and getting enough data to personalize models for individuals is very expensive or even infeasible. This paper addresses these problems by (1) designing a recurrent Generative Adversarial Network to generate realistic synthetic data and to augment the original dataset, (2) enabling the generation of balanced datasets based on heavily unbalanced dataset, and (3) to control the data generation in such a way that the generated data resembles data from specific individuals. We apply these solutions for sleep apnea detection and study in the evaluation the performance of four well-known techniques, i.e., K-Nearest Neighbour, Random Forest, Multi-Layer Perceptron, and Support Vector Machine.

Insights in power grid pixel maps (PGPMs) refer to important facility operating states and unexpected changes in the power grid. Identifying insights helps analysts understand the collaboration of various parts of the grid so that preventive and correct operations can be taken to avoid potential accidents. Existing solutions for identifying insights in PGPMs are performed manually, which may be laborious and expertise-dependent. In this paper, we propose an interactive insight identification and annotation framework by leveraging an enhanced variational autoencoder (VAE). In particular, a new architecture, DenseU-Hierarchical VAE (DUHiV), is designed to learn representations from large-sized PGPMs, which achieves a significantly tighter evidence lower bound (ELBO) than existing Hierarchical VAEs with a Multilayer Perceptron architecture. Our approach supports modulating the derived representations in an interactive visual interface, discover potential insights and create multi-label annotations. Evaluations using real-world PGPMs datasets show that our framework outperforms the baseline models in identifying and annotating insights.

We design a self size-estimating feed-forward network (SSFN) using a joint optimization approach for estimation of number of layers, number of nodes and learning of weight matrices at a low computational complexity. In the proposed approach, SSFN grows from a small-size network to a large-size network. The increase in size from small-size to large-size guarantees a monotonically decreasing cost with addition of nodes and layers. The optimization approach uses a sequence of layer-wise target-seeking non-convex optimization problems. Using `lossless flow property' of some activation functions, such as rectified linear unit (ReLU), we analytically find regularization parameters in the layer-wise non-convex optimization problems. Closed-form analytic expressions of regularization parameters allow to avoid tedious cross-validations. The layer-wise non-convex optimization problems are further relaxed to convex optimization problems for ease of implementation and analytical tractability. The convex relaxation helps to design a low-complexity algorithm for construction of the SSFN. We experiment with eight popular benchmark datasets for sound and image classification tasks. Using extensive experiments we show that the SSFN can self-estimate its size using the low-complexity algorithm. The size of SSFN varies significantly across the eight datasets.

The goal of this paper was to predict the placement in the multiplayer game PUBG (playerunknown battleground). In the game, up to one hundred players parachutes onto an island and scavenge for weapons and equipment to kill others, while avoiding getting killed themselves. The available safe area of the game map decreases in size over time, directing surviving players into tighter areas to force encounters. The last player or team standing wins the round. In this paper specifically, we have tried to predict the placement of the player in the ultimate survival test. The data set has been taken from Kaggle. Entire dataset has 29 attributes which are categories to 1 label(winPlacePerc), training set has 4.5 million instances and testing set has 1.9 million. winPlacePerc is continuous category, which makes it harder to predict the survival of the fittest. To overcome this problem, we have applied multiple machine learning models to find the optimum prediction. Model consists of LightGBM Regression (Light Gradient Boosting Machine Regression), MultiLayer Perceptron, M5P (improvement on C4.5) and Random Forest. To measure the error rate, Mean Absolute Error has been used. With the final prediction we have achieved MAE of 0.02047, 0.065, 0.0592 and 0634 respectively.