A deep learning model can be used to predict a woman's future risk of breast cancer better than traditional practice, according to the findings of a study published in the journal Radiology. The artificial intelligence (AI) model identified women at high risk for breast cancer and placed 31% of all patients with future breast cancer in the top risk decile. This was an improvement over the Tyrer-Cuzick model, which is typically used in clinical practice and was shown to predict risk in just 18% of patients. Using a hybrid model, the researchers found that patients with non-dense breasts and model-assessed high risk had 3.9 times higher cancer incidence than patients with dense breasts and model-assessed low risk. While AI models do not always perform well across diverse races, the hybrid model had an area under the curve of 0.71 for both Caucasian and African American women.

It's a common sentiment that Deep Learning is only good for images and language models. This post is about using Deep Learning on tabular data, for both Regression and Classification problems. We will use fastai library for creating our deep learning models. We will use Kaggle competitions as benchmarks to see how our solutions compares to other solutions using traditional ML models. If you haven't watched fastai tutorials already, please visit this link for the awesome and free tutorials.

Recently, there has been interest in multiplicative recurrent neural networks for language modeling. Indeed, simple Recurrent Neural Networks (RNNs) encounter difficulties recovering from past mistakes when generating sequences due to high correlation between hidden states. These challenges can be mitigated by integrating second-order terms in the hidden-state update. One such model, multiplicative Long Short-Term Memory (mLSTM) is particularly interesting in its original formulation because of the sharing of its second-order term, referred to as the intermediate state. We explore these architectural improvements by introducing new models and testing them on character-level language modeling tasks. This allows us to establish the relevance of shared parametrization in recurrent language modeling.

Deep predictive coding networks are neuroscience-inspired unsupervised learning models that learn to predict future sensory states. We build upon the PredNet implementation by Lotter, Kreiman, and Cox (2016) to investigate if predictive coding representations are useful to predict brain activity in the visual cortex. We use representational similarity analysis (RSA) to compare PredNet representations to functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG) data from the Algonauts Project (Cichy et al., 2019). In contrast to previous findings in the literature (Khaligh-Razavi & Kriegeskorte, 2014), we report empirical data suggesting that unsupervised models trained to predict frames of videos may outperform supervised image classification baselines in terms of correlation to spatial (fMRI) data. Our best submission achieves an average noise normalized correlation score of 16.67% and 27.67% on the fMRI and MEG tracks of the Algonauts Challenge.

A spoken language understanding (SLU) system includes two main tasks, slot filling (SF) and intent detection (ID). The joint model for the two tasks is becoming a tendency in SLU. But the bi-directional interrelated connections between the intent and slots are not established in the existing joint models. In this paper, we propose a novel bi-directional interrelated model for joint intent detection and slot filling. We introduce an SF-ID network to establish direct connections for the two tasks to help them promote each other mutually. Besides, we design an entirely new iteration mechanism inside the SF-ID network to enhance the bi-directional interrelated connections. The experimental results show that the relative improvement in the sentence-level semantic frame accuracy of our model is 3.79% and 5.42% on ATIS and Snips datasets, respectively, compared to the state-of-the-art model.

In this work, we focus on fine-tuning an OpenAI GPT-2 pre-trained model for generating patent claims. GPT-2 has demonstrated impressive efficacy of pre-trained language models on various tasks, particularly coherent text generation. Patent claim language itself has rarely been explored in the past and poses a unique challenge. We are motivated to generate coherent patent claims automatically so that augmented inventing might be viable someday. In our implementation, we identified a unique language structure in patent claims and leveraged its implicit human annotations. We investigated the fine-tuning process by probing the first 100 steps and observing the generated text at each step. Based on both conditional and unconditional random sampling, we analyze the overall quality of generated patent claims. Our contributions include: (1) being the first to generate patent claims by machines and being the first to apply GPT-2 to patent claim generation, (2) providing various experiment results for qualitative analysis and future research, (3) proposing a new sampling approach for text generation, and (4) building an e-mail bot for future researchers to explore the fine-tuned GPT-2 model further.

Modeling the probability distribution of rows in tabular data and generating realistic synthetic data is a non-trivial task. Tabular data usually contains a mix of discrete and continuous columns. Continuous columns may have multiple modes whereas discrete columns are sometimes imbalanced making the modeling difficult. Existing statistical and deep neural network models fail to properly model this type of data. We design TGAN, which uses a conditional generative adversarial network to address these challenges. To aid in a fair and thorough comparison, we design a benchmark with 7 simulated and 8 real datasets and several Bayesian network baselines. TGAN outperforms Bayesian methods on most of the real datasets whereas other deep learning methods could not.

This paper introduces an improved method of wind power prediction via weather forecast-contextualized Long Short- Term Memory Neural Network (LSTM) models. Wind power and weather forecast data were acquired from open-source databases and combined. However, a generic LSTM model performs poorly on this data, with erratic behavior observed on even low-variance data sections. To address this issue, LSTM modifications were proposed and tested for accuracy through both a Normalized Mean Absolute Error and the Naive Ratio, which is a score introduced by this paper to quantify unwanted "naive" model behavior. Results showed an increase in model accuracy with the addition of weather forecast data to the models, as well as major improvements in performance with some model modifications, which are attributed to the increased contextualization and stability of the new models. These new and improved models have the potential to improve power grid stability and expedite renewable power integration.

We address the structure identification and the uniform approximation of two fully nonlinear layer neural networks of the type $f(x)=1^T h(B^T g(A^T x))$ on $\mathbb R^d$ from a small number of query samples. We approach the problem by sampling actively finite difference approximations to Hessians of the network. Gathering several approximate Hessians allows reliably to approximate the matrix subspace $\mathcal W$ spanned by symmetric tensors $a_1 \otimes a_1 ,\dots,a_{m_0}\otimes a_{m_0}$ formed by weights of the first layer together with the entangled symmetric tensors $v_1 \otimes v_1 ,\dots,v_{m_1}\otimes v_{m_1}$, formed by suitable combinations of the weights of the first and second layer as $v_\ell=A G_0 b_\ell/\|A G_0 b_\ell\|_2$, $\ell \in [m_1]$, for a diagonal matrix $G_0$ depending on the activation functions of the first layer. The identification of the 1-rank symmetric tensors within $\mathcal W$ is then performed by the solution of a robust nonlinear program. We provide guarantees of stable recovery under a posteriori verifiable conditions. We further address the correct attribution of approximate weights to the first or second layer. By using a suitably adapted gradient descent iteration, it is possible then to estimate, up to intrinsic symmetries, the shifts of the activations functions of the first layer and compute exactly the matrix $G_0$. Our method of identification of the weights of the network is fully constructive, with quantifiable sample complexity, and therefore contributes to dwindle the black-box nature of the network training phase. We corroborate our theoretical results by extensive numerical experiments.

Intraductal papillary mucinous neoplasm (IPMN) is a precursor to pancreatic ductal adenocarcinoma. While over half of patients are diagnosed with pancreatic cancer at a distant stage, patients who are diagnosed early enjoy a much higher 5-year survival rate of $34\%$ compared to $3\%$ in the former; hence, early diagnosis is key. Unique challenges in the medical imaging domain such as extremely limited annotated data sets and typically large 3D volumetric data have made it difficult for deep learning to secure a strong foothold. In this work, we construct two novel "inflated" deep network architectures, $\textit{InceptINN}$ and $\textit{DenseINN}$, for the task of diagnosing IPMN from multisequence (T1 and T2) MRI. These networks inflate their 2D layers to 3D and bootstrap weights from their 2D counterparts (Inceptionv3 and DenseNet121 respectively) trained on ImageNet to the new 3D kernels. We also extend the inflation process by further expanding the pre-trained kernels to handle any number of input modalities and different fusion strategies. This is one of the first studies to train an end-to-end deep network on multisequence MRI for IPMN diagnosis, and shows that our proposed novel inflated network architectures are able to handle the extremely limited training data (139 MRI scans), while providing an absolute improvement of $8.76\%$ in accuracy for diagnosing IPMN over the current state-of-the-art. Code is publicly available at https://github.com/lalonderodney/INN-Inflated-Neural-Nets.