Artificial intelligence, machine learning, and deep learning all seem like fairly similar concepts. More on the point, they seem the same to an untrained eye. A lot of people see them as the same concept because, in part, they kind of are. Just as nuclear or theoretical physics are parts of physics as a whole, machine learning and deep learning are the branches of a colossal tree that is artificial intelligence. Think of them as Russian nesting dolls, each being a part of something bigger.

Companies have rushed to embrace deep learning's potential in their efforts to automate their enterprises, often with an eye towards replacing as much of their human workforce as possible or to scale their operations without expanding their hiring. An endless stream of success stories tout AI's success in replacing an ever-growing array of traditionally automation-resistant jobs, while developers are hard at work finding ways to replace the rest of them. Yet social media platforms give pause to this idea that deep learning is quite at the inflection point of causing a wave of job displacement. Silicon Valley has been at the vanguard of the AI revolution, pioneering the developments that have pushed the field forward. At the same time, those companies have also been hiring vast armies of human workers to augment the limitations of those very same algorithms.

Compared with all the attention given to the giant AI-related investments made by tech giants, the investments being made by many enterprise software firms don't get as much attention. But as talks with industry execs often drive home, investments in software features that leverage AI/machine learning have also become a key R&D priority for many publicly-traded enterprise software firms -- even if their spending on such work can't compare with what tech giants are doing. When looking at how enterprise software firms are using "AI" to strengthen their offerings, it's worth remembering that there's a difference between broader investments in machine learning, and investments in deep learning in particular. Machine learning (ML) covers the general use of algorithms that analyze data to make conclusions and predictions, and which get better as more data is analyzed. Deep learning (DL) is a computationally demanding subset of machine learning that involves creating artificial neural networks (ANN) that in some respects function like a human brain.

In recent years, deep learning and machine learning have gained traction in so many areas that have a direct positive effect on our lives as well as complex tasks such as computer vision (image recognition), machine translation, and natural language processing. And with like so many other technologies that are changing our lives for good, it has the destructive potential to change it for bad, there is no reason why it won't also be used for malicious activities as well. Up until now, we haven't seen the use of AI for malicious activities in cybersecurity due to the high costs, lack of skills and the tools available. But just like any other technology, it's a matter of time before it happens in cybersecurity. Think about what would happen when attackers start using the power of deep learning and machine learning for their advantage?

Art exhibition "Waterfall of Meaning" by Google PAIR displayed at the Barbican Curve Gallery. The rise of deep learning has been defined by a shift away from transparent and understandable human-written code towards sealed black boxes whose creators have little understanding of how or even why they yield the results they do. Concerns over bias, brittleness and flawed representations have led to growing interest in the area of "explainable AI" in which frameworks help interrogate a model's internal workings to shed light on precisely what it has learned about the world and help its developers nudge it towards a fairer and more faithful internal representation. As companies like Google roll out a growing stable of explainable AI tools like its What-If Tool, perhaps a more transparent and understandable deep learning future can help address the limitations that have slowed the field's deployment. Since the dawn of the computing revolution, the underlying programming that guided those mechanical thinking machines was provided by humans through transparent and visible instruction sets.

The simulated triage workflow let radiologists read scans above the cancer-free threshold, reducing their workload by nearly 20% while also improving their specificity, wrote Adam Yala, with the Massachusetts Institute of Technology in Cambridge, and colleagues.

--Channel charting (CC) has been proposed recently to enable logical positioning of user equipments (UEs) in the neighborhood of a multi-antenna base-station solely from channel-state information (CSI). CC relies on dimensionality reduction of high-dimensional CSI features in order to construct a channel chart that captures spatial and radio geometries so that UEs close in space are close in the channel chart. In this paper, we demonstrate that autoencoder (AE)-based CC can be augmented with side information that is obtained during the CSI acquisition process. More specifically, we propose to include pairwise representation constraints into AEs with the goal of improving the quality of the learned channel charts. We show that such representation-constrained AEs recover the global geometry of the learned channel charts, which enables CC to perform approximate positioning without global navigation satellite systems or supervised learning methods that rely on extensive and expensive measurement campaigns.

Deep learning models have been the subject of study from various perspectives, for example, their training process, interpretation, generalization error, robustness to adversarial attacks, etc. A trained model is defined by its decision boundaries, and therefore, many of the studies about deep learning models speculate about the decision boundaries, and sometimes make simplifying assumptions about them. So far, finding exact points on the decision boundaries of trained deep models has been considered an intractable problem. Here, we compute exact points on the decision boundaries of these models and provide mathematical tools to investigate the surfaces that define the decision boundaries. Through numerical results, we confirm that some of the speculations about the decision boundaries are accurate, some of the computational methods can be improved, and some of the simplifying assumptions may be unreliable, for models with nonlinear activation functions. We advocate for verification of simplifying assumptions and approximation methods, wherever they are used. Finally, we demonstrate that the computational practices used for finding adversarial examples can be improved and computing the closest point on the decision boundary reveals the weakest vulnerability of a model against adversarial attack.

Akshay Deepak Department of Computer Science National Institute of Technology Patna, India Email: akshayd@nitp.ac.in Abstract --Convolutional neural networks (CNN) are widely used for speech emotion recognition (SER). In such cases, the short time fourier transform (STFT) spectrogram is the most popular choice for representing speech, which is fed as input to the CNN. However, the uncertainty principles of the short-time Fourier transform prevent it from capturing time and frequency resolutions simultaneously. On the other hand, the recently proposed single frequency filtering (SFF) spectrogram promises to be a better alternative because it captures both time and frequency resolutions simultaneously. In this work, we explore the SFF spectrogram as an alternative representation of speech for SER. We have modified the SFF spectrogram by taking the average of the amplitudes of all the samples between two successive glottal closure instants (GCI) locations. The duration between two successive GCI locations gives the pitch, motivating us to name the modified SFF spectrogram as pitch-synchronous SFF spectrogram. The GCI locations were detected using zero frequency filtering approach. The proposed pitch-synchronous SFF spectrogram produced accuracy values of 63.95% (unweighted) and 70.4% (weighted) on the IEMOCAP dataset. These correspond to an improvement of 7.35% (unweighted) and 4.3% (weighted) over state-of-the-art result on the STFT sepctrogram using CNN. Specially, the proposed method recognized 22.7% of the happy emotion samples correctly, whereas this number was 0% for state-of-the-art results. These results also promise a much wider use of the proposed pitch-synchronous SFF spectrogram for other speech-based applications. I NTRODUCTION S peech emotion recognition (SER) refers to the classification/recognition of the person's emotional state using the speech signal. SER has a lot of applications in real life.