Search

Models of artificial intelligence (AI) that have billions of parameters can achieve high accuracy across a range of tasks1,2, but they exacerbate the poor energy efficiency of conventional general-purpose processors, such as graphics processing units or central processing units. Analog in-memory computing (analog-AI)3–7 can provide better energy efficiency by performing matrix–vector multiplications in parallel on ‘memory tiles’. However, analog-AI has yet to demonstrate software-equivalent (SWeq) accuracy on models that require many such tiles and efficient communication of neural-network activations between the tiles. Here we present an analog-AI chip that combines 35 million phase-change memory devices across 34 tiles, massively parallel inter-tile communication and analog, low-power peripheral circuitry that can achieve up to 12.4 tera-operations per second per watt (TOPS/W) chip-sustained performance. We demonstrate fully end-to-end SWeq accuracy for a small keyword-spotting network and near-SWeq accuracy on the much larger MLPerf8 recurrent neural-network transducer (RNNT), with more than 45 million weights mapped onto more than 140 million phase-change memory devices across five chips.A low-power chip that runs AI models using analog rather than digital computation shows comparable accuracy on speech-recognition tasks but is more than 14 times as energy efficient. [ABSTRACT FROM AUTHOR]

We discuss inline electrical testing to monitor the baseline of Analog Computing hardware using Phase Change Memory (PCM) technology. Tightening the PCM resistance distribution is necessary to meet analog computation requirement. A new yield methodology is introduced. A study of heater process variation, which will affect the heater height and the PCM resistance, will be discussed. [ABSTRACT FROM AUTHOR]

The impact of bird mortality by collision on windfarms has often been evaluated at the individual level, but rarely at the population level. The Lesser kestrel Falco naumanni is an endangered short‐lived migratory raptor, susceptible to collision with wind turbines. We evaluated the impacts of windfarm turbine collisions on the demography of the largest lesser kestrel population in France. Using data from local monitoring of reproduction and windfarm mortality surveys, combined with capture‐recapture ringing data at a nearby population, we quantified vital parameters of fecundity and survival in order to parameterize a matrix population model to study the viability of this population. The breeding success was high and varied in synchrony with survival probabilities. Between 2013 and 2020, 43 carcasses were found below wind turbines, and when accounting for carcass detection and persistence rates, the true mortality should approach 154 individuals in that period, i.e. 3% of the studied population was affected by collisions each year. The matrix model showed that the population growth observed was only possible if there was a constant recruitment of 26 immigrant individuals each year into the population. Without the excess mortality by the windfarm, we predict that this population would have 22% more breeding pairs than what was observed in 2020. Simulations over 30 years showed that, under the current immigration rate, the population should decline if the excess mortality exceeds 11%. If immigration ceases, the population would decline above 5% excess mortality per year. It is urgent to monitor and reduce the excess mortality by windfarm collisions that threatens this lesser kestrel population. More generally, we advocate the use of population matrix demographic models in impact assessment studies to avoid placing new windfarms close to rare species that could not sustain additional mortality by collisions. [ABSTRACT FROM AUTHOR]

Objectives: The aim of this case series was to collect preliminary data on safety and efficacy of treating cats suffering from refractory feline chronic gingivostomatitis with a single intravenous therapy of cryopreserved placenta‐derived mesenchymal stromal cells. Materials and Methods: We planned the prospective inclusion of cats suffering from refractory chronic gingivostomatitis in three veterinary clinics. All cats received a single infusion of 10×106 cryopreserved cells. Follow‐up evaluations were done at day 15 and at 2‐, 3‐ and 6‐months following infusion. Clinical disease severity was evaluated by dental specialists using a published stomatitis disease activity index scoring system coupled with an owners' assessment questionnaire. Results: All eight cats attended all follow up visits. Cryopreserved ready‐to‐use placenta‐derived cells administered systemically were safe and resulted in notable clinical improvement in all cats as reported by stomatitis disease activity index scoring and owner's survey. Clinical Significance: Infusion of cryopreserved freshly thawed placenta‐derived mesenchymal stromal cells appears to promote clinical and consequently behavioural benefits in cats with refractory chronic gingivostomatitis after having undergone full‐mouth or premolar–molar tooth extraction. Future randomised studies are required to confirm safety and efficacy of this treatment. [ABSTRACT FROM AUTHOR]

Hardware acceleration of deep learning using analog non-volatile memory (NVM) requires large arrays with high device yield, high accuracy Multiply-ACcumulate (MAC) operations, and routing frameworks for implementing arbitrary deep neural network (DNN) topologies. In this article, we present a 14-nm test-chip for Analog AI inference—it contains multiple arrays of phase change memory (PCM)-devices, each array capable of storing 512 $\times $ 512 unique DNN weights and executing massively parallel MAC operations at the location of the data. DNN excitations are transported across the chip using a duration representation on a parallel and reconfigurable 2-D mesh. To accurately transfer inference models to the chip, we describe a closed-loop tuning (CLT) algorithm that programs the four PCM conductances in each weight, achieving <3% average weight-error. A row-wise programming scheme and associated circuitry allow us to execute CLT on up to 512 weights concurrently. We show that the test chip can achieve near-software-equivalent accuracy on two different DNNs. We demonstrate tile-to-tile transport with a fully-on-chip two-layer network for MNIST (accuracy degradation ~0.6%) and show resilience to error propagation across long sequences (up to 10 000 characters) with a recurrent long short-term memory (LSTM) network, implementing off-chip activation and vector-vector operations to generate recurrent inputs used in the next on- chip MAC. [ABSTRACT FROM AUTHOR]

Background: The aim of the present study was to investigate whether microarray gene expression analysis can be used to predict lymph node status in gastric cancer. Methods: Twenty-nine patients undergoing gastrectomy for cancer were enrolled and subdivided according to the pathologic nodal involvement of their disease (N+ vs N0). Molecular profiling was performed by cDNA microarray on tumor tissue and healthy mucosa. Data were processed to identify differently expressed genes. Selected genes were categorized with gene ontology. Results: Compared to healthy gastric mucosa, 52 genes were differently expressed in N+ patients, and 50 genes in N0 patients. Forty-five genes were similarly regulated in N+ and N0 patients, whereas 12 genes were differently expressed between N+ and N0 patients. Seven genes were exclusively expressed in N+ patients: Egr-1 was upregulated; Claudin-18, AKR1C2, Cathepsin E, CA II, TFF 1, and progastricsin were downregulated. Five genes were exclusively expressed in N0 patients: Complement C5 receptor 1, PLA2/VII, and MMP- 9 were upregulated; MAO-A and ID-4 were downregulated. Conclusions: Microarray analysis could be a valuable tool to identify genes associated with lymph node metastasis in gastric cancer. This technique could improve the selection of patients with locally advanced disease who are candidates for extended lymph node dissection, multimodal treatment options, or alternative therapeutic strategies. [ABSTRACT FROM AUTHOR]

Probiotics are described as “friendly bacteria” that could improve the intestine defense by interacting with the resident microflora. There is a large body of evidence suggesting that consumption of functional food containing probiotics exerts positive effects on human health. Several clinical trials have highlighted the efficiency of probiotics in the prevention and treatment of different gastrointestinal disorders including the prevention of antibiotic associated diarrhea, the remission in patients with inflammatory bowel disease, beneficial effects against Helicobacter pylori infection, positive effects in patients affected by allergies and atopic diseases. The clinical benefits of probiotics use are mainly attributed to their antimicrobial substances production and their positive interactions with the enterocytes to reinforce the intestinal epithelial barrier. Moreover, there is evidence suggesting that probiotics stimulate both specific and non-specific host immune responses. Recently, have been published some experiments performed with the DNA microarray technology which provided a global gene screening of the complex bacteria-host interplay. Nevertheless, the molecular mechanisms by which probiotics enhance the intestinal host defense are still not completely elucidated. Here, we review the experiments and clinical studies to date on the complex mechanisms regulating the communication between probiotics and their hosts. [ABSTRACT FROM AUTHOR]