Abstract--Autonomous Vehicles(AV) are one of the brightest promises of the future which would help cut down fatalities and improve travel time while working in harmony. Autonomous vehicles will face with challenging situations and experiences not seen before. These experiences should be converted to knowledge and help the vehicle prepare better in the future. Online Transfer Learning will help transferring prior knowledge to a new task and also keep the knowledge updated as the task evolves. This paper presents the different methods of transfer learning, online transfer learning and organic computing that could be adapted to the domain of autonomous vehicles. Autonomous Vehicles(AV) or Driver-less Cars are one of the most widely discussed emerging technology in the present day. An autonomous vehicle can be explained as a vehicle that adapts to its surroundings and can navigate itself by sensing its environment and minimal human input.

The combination of a CNN detector and a search framework forms the basis for local object/pattern detection. To handle the waste of regional information and the defective compromise between efficiency and accuracy, this paper proposes a probabilistic model with a powerful search framework. By mapping an image into a probabilistic distribution of objects, this new model gives more informative outputs with less computation. The setting and analytic traits are elaborated in this paper, followed by a series of experiments carried out on FDDB, which show that the proposed model is sound, efficient and analytic.

We show that the output of a (residual) convolutional neural network (CNN) with an appropriate prior over the weights and biases is a Gaussian process (GP) in the limit of infinitely many convolutional filters, extending similar results for dense networks. For a CNN, the equivalent kernel can be computed exactly and, unlike "deep kernels", has very few parameters: only the hyperparameters of the original CNN. Further, we show that this kernel has two properties that allow it to be computed efficiently; the cost of evaluating the kernel for a pair of images is similar to a single forward pass through the original CNN with only one filter per layer.

Accurate time-series forecasting is vital for numerous areas of application such as transportation, energy, finance, economics, etc. However, while modern techniques are able to explore large sets of temporal data to build forecasting models, they typically neglect valuable information that is often available under the form of unstructured text. Although this data is in a radically different format, it often contains contextual explanations for many of the patterns that are observed in the temporal data. In this paper, we propose two deep learning architectures that leverage word embeddings, convolutional layers and attention mechanisms for combining text information with time-series data. We apply these approaches for the problem of taxi demand forecasting in event areas. Using publicly available taxi data from New York, we empirically show that by fusing these two complementary cross-modal sources of information, the proposed models are able to significantly reduce the error in the forecasts. Keywords: Deep learning, Data fusion, Cross modality learning, Time series forecasting, Textual data, Taxi demand, Special events, Urban mobility 1. Introduction Understanding what drives the travel behavior of people is a key research topic for developing effective and efficient intelligent transportation systems that adapt to the travel demand. However, typical approaches focus only on capturing recurrent mobility trends that relate to habitual/routine behaviour [1], and on exploiting short-term correlations with recent observation patterns [2, 3]. While this type of approaches can be successful for long-term planning applications or for modeling demand in non-eventful areas such as residential neighborhoods, in lively and highly dynamic areas that are prone to the occurrence of multiple special events, such as music concerts, sports games, festivals, parades and protests, these approaches fail to accurately model mobility demand [4]. As we move towards the deployment of autonomous vehicles, understanding and being able to anticipate mobility demand becomes crucial, especially in shared-mobility scenarios, as this allows for properly managing fleets and increasing user-satisfaction. In order to capture the effects of events, one can exploit the vast amount of information that is shared online about what is planned to take place in the city. However, most of this information is typically in the form of unstructured natural-language text.

The Linear Attention Recurrent Neural Network (LARNN) is a recurrent attention module derived from the Long Short-Term Memory (LSTM) cell and ideas from the consciousness Recurrent Neural Network (RNN). Yes, it LARNNs. The LARNN uses attention on its past cell state values for a limited window size $k$. The formulas are also derived from the Batch Normalized LSTM (BN-LSTM) cell and the Transformer Network for its Multi-Head Attention Mechanism. The Multi-Head Attention Mechanism is used inside the cell such that it can query its own $k$ past values with the attention window. This has the effect of augmenting the rank of the tensor with the attention mechanism, such that the cell can perform complex queries to question its previous inner memories, which should augment the long short-term effect of the memory. With a clever trick, the LARNN cell with attention can be easily used inside a loop on the cell state, just like how any other Recurrent Neural Network (RNN) cell can be looped linearly through time series. This is due to the fact that its state, which is looped upon throughout time steps within time series, stores the inner states in a "first in, first out" queue which contains the $k$ most recent states and on which it is easily possible to add static positional encoding when the queue is represented as a tensor. This neural architecture yields better results than the vanilla LSTM cells. It can obtain results of 91.92% for the test accuracy, compared to the previously attained 91.65% using vanilla LSTM cells. Note that this is not to compare to other research, where up to 93.35% is obtained, but costly using 18 LSTM cells rather than with 2 to 3 cells as analyzed here. Finally, an interesting discovery is made, such that adding activation within the multi-head attention mechanism's linear layers can yield better results in the context researched hereto.

Deep learning has recently been shown to be instrumental in the problem of domain adaptation, where the goal is to learn a model on a target domain using a similar --but not identical-- source domain. The rationale for coupling both techniques is the possibility of extracting common concepts across domains. Considering (strictly) local representations, traditional deep learning assumes common concepts must be captured in the same hidden units. We contend that jointly training a model with source and target data using a single deep network is prone to failure when there is inherently lower-level representational discrepancy between the two domains; such discrepancy leads to a misalignment of corresponding concepts in separate hidden units. We introduce a search framework to correctly align high-level representations when training deep networks; such framework leads to the notion of conceptual --as opposed to representational-- domain adaptation.

Machine learning has become one of the main components for task automation in many application domains. Despite the advancements and impressive achievements of machine learning, it has been shown that learning algorithms can be compromised by attackers both at training and test time. Machine learning systems are especially vulnerable to adversarial examples where small perturbations added to the original data points can produce incorrect or unexpected outputs in the learning algorithms at test time. Mitigation of these attacks is hard as adversarial examples are difficult to detect. Existing related work states that the security of machine learning systems against adversarial examples can be weakened when feature selection is applied to reduce the systems' complexity. In this paper, we empirically disprove this idea, showing that the relative distortion that the attacker has to introduce to succeed in the attack is greater when the target is using a reduced set of features. We also show that the minimal adversarial examples differ statistically more strongly from genuine examples with a lower number of features. However, reducing the feature count can negatively impact the system's performance. We illustrate the trade-off between security and accuracy with specific examples. We propose a design methodology to evaluate the security of machine learning classifiers with embedded feature selection against adversarial examples crafted using different attack strategies.

Stochastic gradient descent (SGD) (Robbins and Monro, 1951) and its variants have been widely used in training deep neural networks. Among those variants, adaptive gradient methods (AdaGrad) (Duchi et al., 2011; McMahan and Streeter, 2010), which scale each coordinate of the gradient by a function of past gradients, can achieve better performance than vanilla SGD in practice when the gradients are sparse. An intuitive explanation for the success of AdaGrad is that it automatically adjusts the learning rate for each feature based on the partial gradient, which accelerates the convergence. However, AdaGrad was later found to demonstrate degraded performance especially in cases where the loss function is nonconvex or the gradient is dense, due to rapid decay of learning rate.

In this paper we address the memory demands that come with the processing of 3-dimensional, high-resolution, multi-channeled medical images in deep learning. We exploit memory-efficient backpropagation techniques, to reduce the memory complexity of network training from being linear in the network's depth, to being roughly constant $ - $ permitting us to elongate deep architectures with negligible memory increase. We evaluate our methodology in the paradigm of Image Quality Transfer, whilst noting its potential application to various tasks that use deep learning. We study the impact of depth on accuracy and show that deeper models have more predictive power, which may exploit larger training sets. We obtain substantially better results than the previous state-of-the-art model with a slight memory increase, reducing the root-mean-squared-error by $ 13\% $. Our code is publicly available.

Short-term load forecasting is one of the crucial sections in smart grid. Precise forecasting enables system operators to make reliable unit commitment and power dispatching decisions. With the advent of big data, a number of artificial intelligence techniques such as back propagation, support vector machine have been used to predict the load of the next day. Nevertheless, due to the noise of raw data and the randomness of power load, forecasting errors of existing approaches are relatively large. In this study, a short-term load forecasting method is proposed on the basis of empirical mode decomposition and long short-term memory networks, the parameters of which are optimized by a particle swarm optimization algorithm. Essentially, empirical mode decomposition can decompose the original time series of historical data into relatively stationary components and long short-term memory network is able to emphasize as well as model the timing of data, the joint use of which is expected to e ffectively apply the characteristics of data itself, so as to improve the predictive accuracy. The e ffectiveness of this research is exemplified on a realistic data set, the experimental results of which show that the proposed method has higher forecasting accuracy and applicability, as compared with existing methods. Introduction Based on historical data, power load forecasting is to explore the developing law of electricity, establish models between power demand and features, then make a valid prediction of future load [1]. A lot of operations in power systems sharply depend on the future information provided by predictions, for example making a satisfying unit commitment (UC) decision [2], saving energy and reducing the cost of power generation [3].